direct democracy and government size: evidence from spain

DIRECT DEMOCRACY AND GOVERNMENT SIZE: EVIDENCE FROM SPAIN

Carlos Sanz

Documentos de Trabajo N.º 1709

2017

DIRECT DEMOCRACY AND GOVERNMENT SIZE: EVIDENCE FROM SPAIN

Documentos de Trabajo. N.º 1709

2017

(*) A previous version of this paper covered a more limited sample period (2002-2011). I am thankful to Manuel Bagüés for sharing the budget data for the entire period in which the 100-inhabitant threshold separated the two systems (1987-2011).(**) E-mail address: [email protected]. Full postal address: C/ Alcalá 48, 28014 Madrid, Spain. I gratefully acknowledge financial support from the Princeton Research Program in Political Economy. I am very thankful to Thomas Fujiwara for his guidance and support. I thank Alfonso Gordaliza for numerous conversations and detailed information on the direct-democracy system, José Miguel Rodríguez for his help in managing and understanding the data, and Samuel Suskind for outstanding research assistance. I also thank Jorge Álvarez, Marco Battaglini, Angus Deaton, Bo Honoré, Federico Huneeus, Matias Iaryczower, John Klopfer, Ilyana Kuziemko, John Londregan, Fernanda Márquez-Padilla, Eduardo Morales, Thomas Romer, Tom Vogl, and seminar participants in Princeton University, the Spanish Economic Association Symposium, Universidad Carlos III de Madrid, Banco de España, Universitat de Barcelona, CaixaBank Research, and the Royal Economic Society Conference for their comments and suggestions. I am solely responsible for any remaining errors.

Carlos Sanz (**)

BANCO DE ESPAÑA

DIRECT DEMOCRACY AND GOVERNMENT SIZE:

EVIDENCE FROM SPAIN (*)

The Working Paper Series seeks to disseminate original research in economics and fi nance. All papers have been anonymously refereed. By publishing these papers, the Banco de España aims to contribute to economic analysis and, in particular, to knowledge of the Spanish economy and its international environment.

The opinions and analyses in the Working Paper Series are the responsibility of the authors and, therefore, do not necessarily coincide with those of the Banco de España or the Eurosystem.

The Banco de España disseminates its main reports and most of its publications via the Internet at the following website: http://www.bde.es.

Reproduction for educational and non-commercial purposes is permitted provided that the source is acknowledged.

© BANCO DE ESPAÑA, Madrid, 2017

ISSN: 1579-8666 (on line)

Abstract

Direct democracy is spreading across the world, but little is known about its effects

on policy. I provide evidence from a unique scenario. In Spain, national law determines

that municipalities follow either direct or representative democracy, depending on their

population. Regression discontinuity estimates indicate that direct democracy leads to

smaller government, reducing public spending by around 8%. Public revenue decreases

by a similar amount and, therefore, there is no effect on budget defi cits. These fi ndings can

be explained by a model in which direct democracy allows voters to enforce lower special-

interest spending.

Keywords: public fi nance, political economy, direct democracy, government spending,

defi cits, budget, regression discontinuity.

JEL classifi cation: D7, H.

Resumen

La democracia directa se está expandiendo por el mundo, pero aún se sabe poco sobre

sus efectos en las políticas públicas. En este artículo se da evidencia a partir de un

escenario único. En España, la Ley determina que los municipios se rijan por un sistema

de democracia directa o representativa, dependiendo de su población. Mediante una

regresión de discontinuidad, se estima que la democracia directa lleva a un menor tamaño

del gobierno, reduciendo el gasto público en alrededor de un 8 %. Los ingresos públicos

disminuyen en una cantidad similar y, por tanto, no hay un efecto en el défi cit. Estos

resultados pueden explicarse mediante un modelo en el que la democracia directa permite

a los votantes imponer un menor gasto en intereses particulares.

Palabras clave: economía pública, economía política, democracia directa, gasto público,

défi cit, presupuestos, regresión de discontinuidad.

Códigos JEL: D7, H.

BANCO DE ESPAÑA 7 DOCUMENTO DE TRABAJO N.º 1709

1 Introduction

A pressing concern about representative democracy is that it leads to a government

that is too large.1 As noted by Besley and Coate (2008), that citizens have only

one vote to cast for candidates who have responsibility for choosing a bundle of

issues may lead to policies that are incongruent with the position favored by the

majority. In particular, if politicians are interested in increasing spending to favor

special-interest groups, then representative democracy will result in overspending.

One way to align policy with the preferences of the median voter is to expand the

use of direct democracy, in which people decide policy initiatives directly.2 By

issue unbundling, direct democracy allows the median voter’s preferences to prevail

along different dimensions and, therefore, constrains politicians’ ability to increase

expenditures (Matsusaka (2005)).

The theoretical development of this argument coincides with a dramatic expan-

sion in the use of direct democracy across the world. In emerging countries, decisions

at the local level are increasingly adopted in participatory meetings. These include

the Gram Panchayat in India, the participatory budget (Conselho do Orcamento

Participativo) in Brazil, and many others in countries such as Bolivia, Indonesia,

Mexico, Peru, Uganda, and Venezuela. In the United States, town meetings are

widely used at the local level in New England, and voters have decided more than

1,600 state-wide ballot propositions in the 21st century.3 In Europe, direct voter

participation is a hallmark of Switzerland and has become increasingly popular at

the local level in Germany. Its use also has been debated in Italy, the Netherlands,

Spain, and in European Union’s institutions.

In this paper, I empirically test whether direct democracy in the form of par-

ticipatory meetings reduces the size of government. Despite the rapid expansion of

direct democracy in recent years, settings with an exogenous source of variation in

the use of direct democracy are rare. As a result, it is difficult to obtain credible

4A noteworthy exception is Hinnerich and Pettersson-Lidbom (2014), but they study a historicalcontext—Sweden in the period following the introduction of universal suffrage in the 1920s—andfocus on welfare spending. Previous literature is summarized by Matsusaka (2005).

causal estimates on the effects of direct democracy.4

1Besley and Case (2003) argue that “there is a widespread belief that agency problems lead toa government that is too large” under representative democracy.

2Direct democracy is a form of democracy in which people decide policy initiatives directly,as opposed to representative democracy, in which people vote for representatives who then de-cide policy initiatives. The main forms of direct democracy are participatory or town meetings,referendums, and initiatives.

3From 2001 to 2015, 1,645 state-wide ballot measures have been voted in the United States.Source: www.ncsl.org. Also in the US, direct democracy has been one of the battlegrounds of theOccupy Wall Street movement.


I provide evidence from a unique scenario to study the effects of direct democ-

racy on policy. In Spain, national law mandates that municipalities operate under a

direct or representative democracy, depending on the size of their population. Mu-

nicipalities with 100 or more inhabitants follow a representative-democracy system,

while those with fewer than 100 inhabitants use direct democracy. Under repre-

sentative democracy, citizens elect a city council every fourth year, and the council

decides on policy. Under direct democracy, the role of the city council is played out

in open town meetings held approximately four times a year.

This institutional framework presents several attractive features. First, the use

of direct or representative democracy is determined by population size, as mandated

by a national law. This allows the use of a regression discontinuity design to esti-

mate effects. Second, no other rule changes at the threshold, unlike what is often the

case for municipal population thresholds. I therefore can attribute the differences

between municipalities at each side of the threshold to the government system and

not to some other regulation. Third, the number of observations is large, leading

to precise estimates. There are more than 8,000 municipalities in Spain, and, fur-

thermore, many municipalities have a population size close to the threshold (around

2,000 municipalities with a population of 250 of fewer inhabitants—see Appendix

Table A1). Fourth, Spanish municipalities have substantial autonomy in both ex-

penditures and revenues. As a result, there is considerable variation in fiscal policy

across them. For example, the municipality in the 90th percentile of expenditures

spends four times more than the one in the 10th percentile.

I employ a rich panel dataset that contains yearly data from the budgets of mu-

nicipalities in the period 1988–2011. An issue that requires careful attention is that

some sorting is observed around the threshold. I show that a regression discontinu-

ity design extended with municipality and year fixed effects, that exploits switches

in government system of municipalities over time, deals well with this issue and

provides credible estimates. I show that there is very good covariate balance around

the threshold, and conduct a battery of robustness checks to assess the validity of

the empirical approach, including testing for pretrends in the outcomes of interests,

separate analyses for switches into and out of direct democracy, placebo tests at


other population thresholds, and donut regressions following Barreca, Guldi, Lindo,

and Waddell (2011), in which observations for which the sorting is concentrated

are dropped. The results from these tests provide assurance of the validity of the

empirical strategy.

The main finding is that direct democracy reduces the size of government. In the

preferred specification, direct democracy reduces spending by almost 8%. Revenues

are decreased by a similar amount, and budget deficits are therefore equal under both

systems. I explain the results with a model, based on Besley and Coate (2008), that

I develop to combine common arguments from two different strands of the literature:

that direct democracy curbs special interest spending, and that it aligns policy with

the preferences of the median voter by issue unbundling. In the model, a majority

of the population (citizens) prefer low special-interest spending, but they care more

for some other dimension of policy. There is, however, a minority (special interests)

that want high spending and whose vote is determined by special-interest spending.

Under representative democracy, the two issues get bundled, as individuals have only

one vote (to elect a representative who will implement policy). The consequence is

that, even though the majority prefer low spending, parties propose high spending in

equilibrium to attract the votes of the special interests. As a result, special-interest

spending is high in representative democracy (independent of which candidate wins

the election). In contrast, in direct democracy, the two dimensions are unbundled, as

individuals vote separately on issues. In this case, the preferences of the median voter

prevail along the two dimensions, and this results in low special-interest spending.

Although the results can be explained by the the mechanism highlighted in this

model, I consider four other potential mechanisms: elite capture in direct democ-

racy, differential costs of participation in meetings and elections, direct democracy

directly affecting what policy individuals prefer, and municipalities in direct democ-

racy obtaining more transfers from upper-level governments. To shed light on their

plausibility, I present five additional sets of results. First, there are no partisan dif-

ferences in spending in representative democracy. This is consistent with the model,

which predicts that the two parties should converge to high spending. Second, ex-

penditures and revenues are equally volatile in direct and representative democracy.

This suggests that the costs of participation in meetings do not drive the results—if

that were the case, policy should be more volatile in direct democracy (Osborne,

Rosenthal, and Turner (2000)). Third, against a hypothesis of gradual learning, the

effects do not seem to change over time: they appear in the first term a municipal-


ity works under direct democracy, and remain similar as municipalities spend more

time under direct democracy. Fourth, direct democracy does not affect subsequent

elections’ behavior (e.g., being exposed to participatory meetings does not lead to

more voter turnout in subsequent elections). Fifth, direct-democracy municipalities

do not obtain more transfers from upper-level governments. In light of this evidence,

and of the political and social characteristics of Spain, I conclude that, although it

is not possible to rule out completely that some of these mechanisms are at work,

the weight of the evidence indicates that they do not drive the results.

This paper contributes to the growing literature on the impact of political in-

stitutions on public policy (see, for example, Besley and Case (2003) and Persson

and Tabellini (2005)). More specifically, it expands our knowledge about the policy

implications of participatory institutions. As mentioned, these institutions are in-

creasingly used in villages across the world, and have been subject of recent work,

both theoretical and empirical. At the theory level, several papers have modeled

decision-making in meetings (Aragones and Sanchez-Pages (2009), Osborne, Rosen-

thal, and Turner (2000), Turner and Weninger (2005)). At the empirical level, the

literature has worked on estimating the impact of participatory institutions on sev-

eral outcomes: for example, Besley, Pande, and Rao (2005) find that village meetings

in South India improved the targeting of welfare programs for the poor; Hinnerich

and Pettersson-Lidbom (2014) find that town meetings reduced welfare spending

in the context of Sweden in the 1920s and 30s; and Olken (2010), in a randomized

controlled trial, finds that the use of plebiscites improved voter satisfaction with

public policy in Indonesian villages. The present paper adds to that literature by

providing the first quasi-experimental evidence on how town meetings affect public

spending and taxation decisions in a contemporary setting.

This paper also contributes to the literature, pioneered by De Haan and Sturm

(1997) and Roubini and Sachs (1989), that analyzes how public budgets are shaped

by political institutions, and, in particular, by direct-democracy institutions. Evi-

dence from the US and Switzerland suggests that initiative and referendums reduce

public spending (see Matsusaka (1995) for the US, and Feld and Matsusaka (2003)

and Funk and Gathmann (2011) for Switzerland). Berry (2014), however, has re-

cently questioned the causal interpretation of the findings for the US, and Asatryan,

Baskaran, Grigoriadis, and Heinemann (2015) and Asatryan (2016) have found that

direct democracy at the local level increased expenditures in Germany. This pa-

per contributes to that literature by providing evidence from an unexplored setting,


in which, unlike what happens in the aforementioned countries, the use of direct

democracy is determined by population size, therefore reducing endogeneity con-

cerns.

Finally, this paper expands the work that uses regression discontinuity designs

to estimate the effects of institutional rules on policy (for example, Bordignon, Nan-

nicini, and Tabellini (2016), Ferraz and Finan (2009), Fujiwara (2015), Gagliarducci,

Nannicini, and Naticchioni (2011), Gagliarducci and Nannicini (2013), Hinnerich

and Pettersson-Lidbom (2014), and Pettersson-Lidbom (2012)).

The rest of the paper is organized as follows. Section 2 presents the theoretical

framework. Section 3 provides background on the Spanish government systems. The

empirical strategy and the data are outlined in Sections 4 and 5. Section 6 presents

the main results. The robustness of the results is the focus of Section 7. Section 8

explores other potential mechanisms. Section 9 concludes the paper.

2 Theoretical Framework

Social scientists have long been interested in how switching from representative to

direct democracy might change policy. One strand of the literature argues that, in

representative democracy, an agency problem exists between voters and their elected

representatives due to free-rider problems in monitoring and disciplining officehold-

ers (Matsusaka (2005)). This gives elected officials leeway to pursue costly policies

that are not in the interests of their constituents, resulting in overspending. Frey

(1994) argues that a model that pictures politicians as forming a coalition against

taxpayers and voters seems to be an apt illustration of representative democracy. In

this view, the elected representatives are a well-defined group that jointly reaps rents

or cartelizes against the interests of citizens. Besley and Case (2003), in summariz-

ing this literature, conclude that “there is a widespread belief that agency problems

lead to a government that is too large” under representative democracy. The use of

direct-democracy institutions is a way to address this issue. By allowing people to

vote directly on policy, direct democracy gives them a way to circumvent represen-

tative institutions that may have been captured by elites or other special interests

and aligns policy with the preferences of the median voter (Matsusaka (2005)).

Another strand of the literature emphasizes that, under representative democ-

racy, citizens have only one vote to cast for candidates who have responsibility for

choosing a bundle of issues (Besley and Coate (2008)). This may lead to policies that


are incongruent with the position favored by the median voter in some of the dimen-

sions. By unbundling issues, direct democracy allows the median voter’s preferences

to prevail along different dimensions.

I formalize these ideas in a simple model, adapted from Besley and Coate (2008),

that I summarize here and present in full in the Appendix. In the model, policy is

two-dimensional. There is a main ideological issue (henceforth, ideology) and a sec-

ondary issue (special-interest spending). Although the majority of individuals prefer

low (possibly zero) special-interest spending, there exists a minority that prefer high

spending, perhaps because they will benefit directly from it. The majority care more

about ideology but the minority care more about spending. Under representative

democracy, policy is implemented by a representative elected in an election between

two candidates, proposed by two political parties. Assuming the minority is suffi-

ciently large, in the unique equilibrium, both parties propose high special-interest

spending, against the wishes of the majority. In contrast, under direct democracy,

individuals vote separately on the ideological issue and on special-interest spending,

and the position favored by the median voter in each dimension is chosen. Direct

democracy, therefore, results in low special-interest spending, unlike what happens

under representative democracy.

The model, therefore, blends the points made by two different strands of the

literature: that direct democracy curbs special interest spending, and that it aligns

policy with the preferences of the median voter through the unbundling of issues.

In addition, the model predicts that spending in representative democracy should

not depend on which party is in office. In Section 6, I provide empirical evidence

that supports this additional prediction, and discuss other alternative mechanisms.

3 Institutional Background

In this section, I provide some institutional background and describe the two gov-

ernment systems that are used by Spanish municipalities. Spain is a very decen-

tralized country. It is politically divided into 17 regions, 50 provinces, and more

than 8,000 municipalities, which are administered by local governments. Local gov-

ernments spend 13% of the overall spending of the country (Gil-Ruiz Gil-Esparza

and Iglesias Quintana (2007)) and have substantial autonomy. National law (Local

Government Regulatory Law, Ley de Bases del Regimen Local) requires them to

provide a variety of services, including public lighting, waste collection, street clean-


ing, road paving, household water supply, sewerage, access to villages, and food and

beverage control. In addition, they usually provide other services useful to the mu-

nicipality, such as organizing local festivities or providing tourist information. Local

governments can decide how much to spend on the required goods and services and

whether to provide additional goods and services. There is considerable variation

in expenditures across municipalities: The 90–10 ratio is 4.6.5

Local governments obtain their own revenues from taxes and fees: they can set

the tax rates—within certain limits imposed by national law—and the prices and

user fees of the goods and services that they provide. According to national law,

municipalities can impose five different taxes: property tax (impuesto de bienes

inmuebles), a tax on economic activities (impuesto de actividades economicas), a

motor vehicle tax (impuesto sobre los vehıculos de traccion mecanica), a tax on

improvements to real property (impuesto de instalaciones, construcciones y obras),

and a tax on the increased value of urban land (impuesto sobre el incremento de

valor de los terrenos de naturaleza urbana).6 As happens with expenditures, there

is substantial variation in revenues across municipalities: The 90–10 ratio is 4.5.

In addition to these own expenditures and revenues, some expenditures are fi-

nanced through transfers from upper-level governments.7 Throughout the paper, I

focus on the municipalities’ own expenditures and revenues (that is, overall revenues

and expenditures net of transfers), as they are the ones that can be controlled by

local governments. In Section 8, I study if the amount of received transfers can drive

the results of the paper.

The Law on Budget Stability (Ley de Estabilidad Presupuestaria) states that

municipalities must have a balanced budget.8 However, no penalty is stipulated for

municipalities that fail to comply with the law. Indeed, there was a deficit in 39%

of the budgets in the sample, and in many cases a sizable one: for example, in 14%

of the budgets I observe a deficit of more than 3% of revenues (see the histogram of

5I calculate this number by taking the average value of expenditures by municipality over timefor those municipalities. This way, I do not use the cross-time variation, which would overstate thevariation in expenditures.

6 The first three are required taxes that national law forces municipalities to levy while the lasttwo are optional taxes. For both required and optional taxes, municipalities are free to set the taxrates and, in some cases, introduce exemptions and deductions, within certain limits imposed bynational law. For example, for the most important tax, the property tax, they are allowed to setthe tax rate on urban land from 0.4 to 1.1% of the property value.

7Transfers account for approximately 46% of overall expenditures and revenues.8In 2011, in the midst of the European debt crisis, the Constitution was amended to include

this principle.


deficits in Appendix Figure A2).9 For this reason, it makes sense to test whether

direct democracy has an effect on deficits.

Municipalities must follow a government system that is determined by their pop-

ulation size one year before the local elections, which are held simultaneously in all

municipalities in the country every four years. Therefore, municipalities change the

government system at most once every four years. During the sample period (1988–

2011), municipalities with 100 or more inhabitants had to followed a representative-

democracy system, while those with fewer than 100 inhabitants followed a direct-

democracy system (called “open council”, or concejo abierto). Appendix Table A1

shows the number of direct-democracy municipalities covered in the sample, by

term. Between 603 and 827 municipalities operated under direct democracy, with

the number increasing over time as a consequence of the declining population of

small municipalities.

In the representative-democracy system, individuals elect a city council every

four years on local election day. The council elects a mayor among its members and

is entitled to approve the budget, decide on expenditures in various fields, control

the governing bodies, and to decide on the roll-call vote of confidence on the mayor.

The mayor chairs the meetings of the council, casts the decisive vote in the event of

a tie, heads the local police, and appoints mayoral deputies and cabinet members,

among other responsibilities. The size of the city council and the electoral rules

differ according to the population of the municipality. In particular, there is an

important change in the electoral system at 250 inhabitants.10 Municipalities below

that population size elect a council of five members in an open-list, plurality-at-large

election.11 To avoid dealing with multiple thresholds, and given that the estimation

will exploit municipalities with a population size close to the 100-inhabitant thresh-

old, throughout the paper I restrict the sample to municipalities with 250 or fewer

inhabitants.12

9Deficits can be financed by issuing debt or by reducing cash balances. The Law on Local Gov-ernment Financing (Ley de Haciendas Locales) authorizes municipalities to issue debt, but withsome limits. For example, if the debt outstanding exceeds 110% of current revenues in the pre-vious year, municipalities must ask for the authorization of the regional or national governmentsand present a consolidation plan (see articles 48-55 of the (Ley de Haciendas Locales) for a com-plete description of the requirements). However, that there was no penalty associated with thoseregulations led many municipalities to not observe the law (Vila (2012)).

10Sanz (2015) studies the effects of that discontinuity on voter turnout.11More specifically, political parties can present candidate lists of up to five candidates, and

voters can vote for up to four candidates from the same or different party lists. The five most-votedcandidates are elected members of the council.

12Also, I drop from the analysis the nine newly created municipalities as well as the municipalitiesfrom which they seceded.


In the direct-democracy system, the role of the city council is played through

open town meetings, so most relevant decisions, including the budget, must be

approved directly by individuals in a pure direct democracy framework.13 (In Ap-

pendix B, I describe the origins and the evolution in the application of this system.)

The specific working rules of the meetings may differ across municipalities, as the

law states that municipalities can work according to the local traditions. However,

national law (Real Decreto 2568/1986) provides some guidelines. Any individual el-

igible to vote (age 18+) may attend the meetings. The meetings are chaired by the

mayor, who is elected by first-past-the-post on local election day. Town meetings are

to be called on a Sunday or a holiday at least once every three months. A quorum

of one-third of the population is required. Citizens can delegate their vote to any

other citizen of the municipality, but no citizen can represent more than a third of

all individuals.14 Decisions are adopted by simple plurality rule. Data for atten-

dance are not publicly available, but I have collected data by hand from the minutes

of the meetings in one Spanish province.15 Average attendance in the sample was

11.6%. Taking vote delegation into account, the average share of represented voters

was 43.7%.

4 Data

Data for the public budgets and population size of municipalities are from the Span-

ish Ministry of Finance (Ministerio de Hacienda) and are publicly available for

73.6% of annual budgets.16 During the sample period, local elections were held

every four years from 1987 to 2011, so the data set covers six terms. Because local

elections are in May or June, the (possible) change in government system happens

13Figure 1 shows a meeting in the municipality of Madarcos, in the region of Madrid.14The delegation of the vote must be written and can be for a specific meeting or permanent.15The minutes of the meetings are stored in the archives of the regional administration. I collected

data from the province of Valladolid in the region of Castilla y Leon from the years 2006 and 2008.Data on attendance were available for 31% (11/36) of municipalities, as attendance at the meetingsis recorded only if the secretary wishes to do so.

16Municipalities were required to report their budgets to the Ministry of Finance, which makesthem public. However, data from some budgets are missing because either they did not comply withtheir obligations or because their data have not been digitized. A placebo test shows that directdemocracy does not have an effect on a dummy variable that indicates whether the observation isavailable (see Appendix Table A8).


halfway through the year. As spending and revenues are determined by the bud-

get approved at the end of the previous year, I assign election-year observations to

the previous term.17 Appedix Figure A1 provides a graphical representation of this

information.18

Table 1 shows the summary statistics. The first three variables are the main

outcomes: real expenditures per capita (Expenditures), real revenues per capita

(Revenues), and budget deficit (Deficit). To obtain the variables in real terms, I

divide the nominal variables by the GDP deflator.19 The deficit is the difference

between expenditures and revenues. The average municipality collects e647 per

capita per year in taxes and fees, and spends e671. Therefore, the average deficit

is e24 per capita per year. Appendix Figure A2 shows the histograms for these

variables. The next four variables, obtained from the Spanish Ministry of Interior

(Ministerio del Interior), are used to study the relevance of political parties in

driving the results. RW Mayor and LW Mayor are dummy variables that indicate

whether the mayor is from the main right-wing party (Popular Party, PP) or the

main left-wing party (Socialist Party, PSOE). With the sample restricted to those

elections in which these two were the two most-voted parties, the next variables

are the percent-point difference between the PP and the PSOE RW Difference,

and a dummy that indicates whether the PP was the most voted (so that a positive

(negative) value of RW Difference implies RW Winner = 1 (RW Winner = 0)). The

next six variables, provided by the National Statistics Institute (Instituto Nacional

de Estadıstica), are from national Congress electoral results, and are used to test

17In Spain, the fiscal year coincides with the calendar year, and budgets for year y are approvedin November or December of year y − 1.

18If expenditures and revenues are sticky, expenditures and revenues for municipality m at year yare partially determined by the government system that municipality m followed in years before y.When a municipality m switches systems, this implies that some of the expenditures and revenuesfor year y may be determined by a government system that does not correspond to the governmentsystem municipality m is following at that year. In general, this will make the estimated coefficientscloser to zero than in a scenario with no changes in government system. These considerations,however, are not likely to play a significant role in practice. First, most expenditures and revenuesin small municipalities are decided on a year-to-year basis. For example, it is rare that theyincur in multi-year capital expenditures, unlike what happens in larger municipalities. It is true,nonetheless, that other expenditures, such as personnel, could be more sticky. Second, if stickypolicies were playing a large role in the estimation, we should expect to see that the effects of directdemocracy on policy increase (in absolute value) over time, as municipalities spend more time in agiven system. However, I find that the effects of direct democracy on policy are already producedin the first year that a municipality switches into direct democracy and do not vary significantlyover time (see Section 8).

19Data for the deflator are from the Spanish National Institute for Statistics.


for covariate balance around the threshold. They are the shares of votes for the

three main parties in Spain—the right-wing PP (Votes Right), the left-wing PSOE

(Votes Left), and the far-left-wing United Left (Votes Far Left)—, the difference in

the share of votes for the two most-voted parties (Votes Difference), the percentage

of votes for the most-voted party (Votes Winner), and voter turnout (Turnout).20

The final six variables are demographic variables, which I also employ to study

covariate smoothness around the threshold, are the average age in the municipality

(Mean Age); the share of young (Young), middle-aged (Middle-Aged), and old (Old)

individuals; the share of immigrants (Immigrants); and the share of EU immigrants

over total immigrants (EU Immigrants).21

5 Empirical Strategy

A discontinuity in the density of population sizes is observed at the threshold (see

Figure 2).22 The shape of the discontinuity raises the concern that some munic-

ipalities self-select into representative democracy, and that, as a result, direct-

and representative-democracy municipalities differ in other characteristics that may

themselves affect policy. I take this issue seriously. First, I will discuss why and

how the sorting around the threshold arises, and then I will explain how I deal with

this issue in the empirical approach.23

In Spain, the official population size of a municipality is given by the number

of citizens who are registered in the municipal registry (padron municipal). Mu-

nicipalities keep track of all the variations in the population in the public registry

and report periodically the data to the National Statistics Institute (INE). The INE

validates the information it receives, checking that there is no fraud—for example,

20Some election variable files are missing from the official files, and, for this reason, the numberof observations is lower than for the budget and population variables. In addition, the sample sizeis smaller for Votes Far Left because the United Left did not run in some regions in some elections.See Appendix C for a precise definition of the election variables.

21The precise definition of these variables is in Appendix C. They are available only from 1996.Also note that, in municipalities with no immigrants, the share of EU immigrants is missing byconstruction. For this reason, the number of observations is lower for EU Immigrants.

22The discontinuity is significant at the 1% level, according to McCrary (2008)’s test.23Sorting in regression discontinuity designs that use population thresholds is studied in depth

by Eggers, Freier, Grembi, and Nannicini (2015), who provide evidence that sorting also exists inFrance, Germany, and Italy. They also argue that, even when extensive strategic sorting is takingplace, a regression discontinuity design “may still be an attractive approach to studying policyeffects; after all, the alternative is to study settings where all political units (not just those that areable to manipulate their population figures) choose which policies to adopt”.


it ensures that, for every registration, there is a corresponding unregistration in an-

other municipality—and, yearly, makes the final population figures public. While

this system makes it difficult to imagine that there is direct fraud or manipulation

of population figures, sorting around the threshold can appear as mayors (or other

local politicians) persuade some individuals to register in the municipality, with the

goal of reaching the population threshold and falling into representative democracy.

This is facilitated by the fact that individuals who have dwellings in more than one

municipality can, in practice, decide in which of them to register: Although individ-

uals are required to register in the municipality in which they spend the most time,

this requirement is almost impossible to monitor and is not enforced in practice.

Naturally, the question is why politicians would prefer to be under representative

democracy. There are at least three possible reasons. One is that representative

democracy is easier to operate, as it does not require calling town meetings to adopt

decisions.24 A second possible reason is that there are five political positions in

representative democracy (the five city councilors), but only one in direct democracy

(the mayor). Although, in most cases, these positions are not remunerated, people

may still derive non-monetary benefits for holding them. Finally, following the logic

of the model presented in Section 2, if the mayor is a special interest, he or she will

prefer to be under representative democracy as a means to implement his or her

preferred policy.25

While the existence of sorting is certainly an issue that requires close attention, it

does not imply, by itself, that the estimates will be biased. In his canonical work on

sorting, McCrary (2008) explains that there may be a discontinuity in the running

variable even when there is no failure of identification, so that a continuous density

is not necessary (or sufficient) for identification, except under auxiliary assumptions.

24Conversations with mayors and other local government officials make me think that this isprobably the most relevant reason.

25Of course, if the mayor is a citizen instead of a special interest, he or she will prefer directdemocracy. This raises the theoretical possibility of two-way sorting. Although it is not possibleto directly test for the existence of two-way sorting, I believe that sorting into direct democracyis negligible. First, as mentioned, mayors and local government officials indicate that the mainreason for sorting is the first, i.e., that representative democracy is easier to operate. Second, it ishard to conceive local government officials trying to persuade people to go to register in anothermunicipality. Also, note that the empirical approach and the robustness checks to assess the validityof the strategy, in particular the donut regressions, do not depend on the sorting being one-way.


Outcomemyt = βDirDemmt + f(Populationmt − 100) + umyt, (1)

where Outcomemyt is the outcome of interest in municipality m at year y in term t,

DirDemmt is the treatment dummy variable that takes the value of 1 if municipality

m follows direct democracy in term t and 0 otherwise, Populationmt is the assignment

variable (population one year before the local elections), f is a smooth function of

the assignment variable, and umyt is an error term. The parameter of interest is β.

The results from estimating Equation (1) show a reduction of expenditures and

revenues in direct democracy (see Appendix Table A2). However, there are some

concerns about the validity of this approach. In particular, there is some covariate

imbalance (Table A3) and evidence of a pretrend in the outcomes of interest (Table

A4).

Next, to deal with these issues, I consider a regression discontinuity design

extended with fixed effects. I find that this strategy yields credible estimates of

the treatment effects, and use it as the main specification throughout the paper.26

Specifically, I consider the following estimating equation:

Outcomemyt = αm + γy + βDirDemmt + f(Populationmt − 100) + umyt, (2)

where αm is a municipality fixed effect and γy is a year fixed effect.

Identification is therefore based on switches in government system, of which there

are 513 in the sample (see Appendix Table A1).27 The identification assumption

is that the characteristics that make a municipality more likely to sort across the

threshold are not correlated with other characteristics that affect policy, conditional

on the fixed effects. This implies that it is the timing of the sorting that matters.

26An empirical strategy that combines regression discontinuity and fixed effects is also followedby Pettersson-Lidbom (2012) and Corbi, Papaioannou, and Surico (2014). Other papers that userelated strategies that combine the cross-sectional discontinuity and the longitudinal structure ofthe data are ?) Campa (2011), Casas-Arce and Saiz (2015), Cellini, Ferreira, and Rothstein (2010)Deshpande (2014), Grembi, Nannicini, and Troiano (2012), Lemieux and Milligan (2008).

27Most of these switches (379) are from representative to direct democracy, as there is a generaltrend of a falling population in small Spanish municipalities. As a result, the empirical strategyis based mainly on these switches instead of on switches out of direct democracy, which are moresuspicious, given the shape of the sorting. In Section 7, I show the results of the estimation basedon considering the two types of switches separately.

I start by considering a standard regression-discontinuity design framework:


For example, suppose that the whole sorting process is driven by local government

officials who are very good at persuading individuals to register. A violation of the

identification assumption would not be that local government officials’ characteris-

tics are correlated with other factors that affect policy but, rather, that the precise

period in which the sorting is achieved coincides with a change in those other factors.

Although the identification assumption is, of course, not directly testable, I

perform five sets of tests to assess the validity of this empirical approach, following

previous research that has dealt with similar situations (Pettersson-Lidbom (2012),

Barreca, Guldi, Lindo, and Waddell (2011)). First, I check if (conditional on the

fixed effects) municipalities at each side of the threshold are similar in other time-

varying variables that may have an effect on the outcomes of interest. Second, I test

if there is an effect on the outcomes at the previous period, and if the results are

robust to including municipality-specific trends. Third, I test whether the effect is

similar in municipalities that switch into direct democracy and in those that switch

out of direct democracy. Finding that the effect is similar would reinforce the

credibility of the estimates, as it would be difficult to explain that correlation from

some omitted factor. Fourth, I conduct placebo tests at other (artificial) population

thresholds. Fifth, I consider donut regressions, following Barreca, Guldi, Lindo, and

Waddell (2011). In this test, observations very close to the threshold, where self-

selection is concentrated, are dropped from the analysis. The results from all these

tests provide assurance of the validity of the empirical strategy.

I use nonparametric local linear regressions to estimate Equation (2), as sug-

gested by Hahn and der Klaauw (2001), Porter (2003) and Gelman and Imbens

(2014). A key ingredient to this approach is the bandwidth. A larger bandwidth

increases the efficiency of the estimation at the cost of more bias. I choose a base-

line bandwidth according to the procedure suggested by Imbens and Kalyanaraman

(2012) and provide the results at different fractions of that bandwidth.28 I use a

rectangular kernel, as recommended by Imbens and Lemieux (2008) and Lee and

Lemieux (2010). This is equivalent to estimating standard linear regressions over

the interval of the selected bandwidth on both sides of the cutoff point. I clus-

ter standard errors at the municipality level (Bertrand, Duflo, and Mullainathan

(2004)). Because the running variable is discrete, I also cluster standard errors at

the running variable, as recommended by Lee and Card (2008). Therefore, I follow

28The results are very similar if I use the optimal bandwidth suggested by Calonico, Cattaneo,and Titiunik (2014). This method selects slightly smaller bandwidths.

a multi-clustering approach (Cameron, Gelbach, and Miller (2011)).29

29The results are very similar if standard errors are clustered just by municipality.


6 Main Results: Effect of Direct Democracy on the Size

of Government

This section presents the main results of the paper. Additional results will be dis-

cussed in Section 8 to shed light on the mechanisms. Table 2 presents the estimates

of the impact of direct democracy on the public finances of local governments. The

table shows the results of estimating Equation (2), with the log of Expenditures,

the log of Revenues, and Deficit as the outcomes. A graphical representation of the

results is shown in Figure 3.30

I find evidence of an effect of direct democracy on public expenditures (Panel A

of Table 2). Under the preferred specification (Column 1), which uses the bandwidth

from Imbens and Kalyanaraman (2012), direct democracy reduces public spending

by almost 8%, and the effect is significant at the 1% level. On average, municipalities

just to the right of the threshold (100 inhabitants) spend e709 per capita; thus, the

effect is equivalent to a decrease of e57 per capita.31 Next, I study robustness to

bandwidth choice and alternative specifications. Columns 2 and 3 show the results

for 50% and 150% of the optimal bandwidth (Appendix Figure A3 displays the

results graphically for a wide range of bandwidths). Columns 4 to 6 show that the

results are robust to fitting high-order polynomials to the whole sample. Finally, as

shown in Column 7, the results are also robust to an alternative specification that

includes municipality-specific time trends. This specification will be discussed in

more detail in the next section.

Panels B and C of Table 2 concern whether the differences in public spending go

together with a change in revenues or whether they are created by different deficits.

According to the preferred specification (Column 2), direct democracy reduces rev-

30To present a graphical representation that incorporates the fixed effects and, therefore, matches

the estimated equation, I estimate Outcomemyt = αm + γy +100+OBW∑

j=100−OBW

δjPopulationj ,mt) + umyt

using the observations within the optimal bandwidth for Outcomemyt , where Populationj ,mt is adummy that indicates whether municipality m has population size j at term t. Then, in the y-axis,I plot the estimated coefficients δj . I normalize the coefficients so that the average bin immediatelyto the right of the threshold takes the value of zero.

31Because the outcome variable is the log of expenditures, the exact percent effect on expendituresis 100 ∗ (exp(−0.0377)− 1) = −7.688 (Halvorsen and Palmquist (1980)).


enues by 5%, and the effect is statistically significant at the 1% level. The rest of

the columns and Figure A3 show that the results are robust to different bandwidths

and specifications. Finally, the results for deficits (Panel C) are close to zero and

not significant under any specification. Therefore, the evidence indicates that di-

rect democracy reduces expenditures and revenues by a similar amount, without

affecting budget deficits.

7 Robustness

In this section, I present five robustness checks to assess the validity of the empirical

strategy: placebo tests to study covariate balance at the threshold, placebo tests for

pretrends in the outcomes of interest, placebo tests at other (artificial) population

thresholds, separate analyses for municipalities that switch into and out of direct

democracy, and donut regressions. In the Appendix, I show additional robustness

checks to study whether the results change when top-coding outliers (Table A7) and

whether the number of missing observations changes at the threshold (Table A8).

7.1 Covariate Smoothness

I study covariate smoothness around the threshold by performing placebo tests:

I estimate Equation (2) with a variety of political and demographic variables as

outcomes. If the empirical strategy is valid, there should be no effect of direct

democracy on these variables.

A possible concern is that municipalities that switch into direct democracy do a

more conservative policy only because they are becoming more ideologically conser-

vative. To test for this, I exploit data from national Congress elections. These data

are especially suited for this purpose because, although results are reported at the

municipality level, the electoral district is at the province level and, therefore, the

100-inhabitant threshold does not play any role in the elections. Further, Congress

elections are the most important elections in Spain, and turnout is high (78% in

the average municipality during the sample period); thus, they are likely to capture

political differences across municipalities.32 I estimate Equation (2) with previous

elections’ variables as the outcomes. For year y in term t, I consider Congress elec-

tions held during the previous term t− 1. In particular, I use the share of votes for

32Spain is a parliamentary democracy, so there are no direct elections for the executive branch.Citizens elect the Congress, which, in turn, elects the Prime Minister.


the three main political parties—the right-wing PP (Votes Rightm,t−1), the left-wingPSOE (Votes Leftm,t−1), and the far-left-wing United Left (Votes Far Leftm,t−1)—,

the difference in the share of votes for the two most-voted parties and the share

of votes for the most-voted party (Votes Differencem,t−1, and Votes Winnerm,t−1),which give a measure of how politically polarized the municipality is, and voter

turnout (Turnoutm,t−1), which has been shown to correlate with variables that also

may affect people’s preferences toward economic policy, such as age, education, or

income (Blais and Achen (2010), Matsusaka and Palda (1999)).33 Next, I study co-

variate smoothness in demographic variables: average age in the municipality (Mean

Age); the share of young (Young), middle-aged (Middle-Aged), and old (Old) in-

dividuals; the share of immigrants (Immigrants); and the share of EU immigrants

over total immigrants (EU Immigrants). These variables will capture how economic

conditions are evolving, as immigrants are more likely to locate in booming munic-

ipalities.

The results from these tests are shown Table 3 and represented graphically in

Figure 4. All of the variables are smooth at the threshold, with no coefficient being

significant at even the 10% level. This provides assurance about the validity of the

empirical strategy, as it indicates that, conditional on the fixed effects, municipalities

at both sides of the threshold are similar in a variety of political and demographic

variables.

7.2 Pretrends in the Outcomes of Interest

In this section, I exploit the longitudinal dimension of the data to test for pretrends

in the outcomes of interest. First, I lag the outcome variables four years (so that

they correspond to the outcomes at the same year of the previous term) and perform

placebo tests by estimating Equation (2) with these lagged variables as outcomes.

Finding a discontinuity in the contemporaneous outcome but not in the lagged

one would be a strong piece of evidence supporting the validity of the regression

discontinuity design (Lee and Lemieux (2010)). The results of these tests are shown

in Table 4 and Appendix Figure A4. The coefficients in Table 4 are starkly different

to those that show the contemporaneous effect shown in Table 2, and show that

there is no effect of direct democracy on the previous term’s expenditures, revenues

and deficit.

33The rest of the parties with representation in Congress during the entire period are regionalparties that do not run in the whole country.


An alternative check on the identification strategy is to add municipality-specific

time trends to the main specification. In particular, I consider:

Outcomemyt = α0m + α1mt + γy +˜βDirDemmt + f(Populationmt − 100) + ˜umyt, (3)

where α0m is a municipality-specific intercept, as before, and α1mt is a municipality

coefficient multiplying the time-trend variable, y. The last column of Table 2 shows

that the results are robust to this alternative specification, with direct democracy

reducing expenditures and revenues by approximately 5% (results significant at the

10% and 5% levels, respectively).

The results in this section appease the concern that municipalities that switch

systems in a given direction were conducting different policies than the rest before

the switch. In particular, one concern would be that mayors would try to cross the

threshold by using taxes or public spending to attract people to the municipality.

If this happened, we should see an effect of direct democracy this term on taxes or

spending in the previous term. The results rule out this possibility.

7.3 Switches into and out of Direct Democracy

The preferred specification is based on switches in the government system. Given

the sorting process and the shape of the density of population sizes, which suggests a

preference for representative democracy, one concern is that the results are driven by

time-varying omitted variables associated with a switch from direct to representative

democracy. I study whether this is the case by performing two separate regressions.

I first base the estimation on switches into direct democracy and exclude from the

sample the municipality-years that correspond to terms in which a municipality

switched out of direct into representative democracy. I then estimate Equation

(2) with the restricted sample. Next, I do the reverse and drop the municipality-

years that correspond to switches out of direct into representative democracy. If

the results from these two different samples point to a reduction in expenditures

and revenues, that would reinforce the credibility of the estimates, as it would be

difficult to explain that correlation by some omitted factor.

Appendix Table A5 shows the results. For expenditures, the preferred specifica-

tion for switches into direct democracy yields a point estimate of -6.4% (significant

at the 5% level), and the one for switches out of direct democracy, -13.1% (sig-

nificant at the 1% level). The results for revenues also tend to support the main


findings. In switches into direct democracy, direct democracy reduces revenues by

4.2% in the preferred specification—the effect is not significant, but it is significant

at the 5% level at both 50% and 150% of that bandwidth. In switches out of direct

democracy, the effect is -8.4% (significant at 10%). Finally, the results for deficits

are mostly insignificant in the two different subsamples.

7.4 Placebo Tests at Other Population Thresholds

In this section I conduct placebo tests by estimating the effect of crossing population

thresholds that are irrelevant (e.g., the effect of having more than 115 inhabitants).

Specifically, I create placebo treatments at all other population sizes from 30 to 220

inhabitants, by defining dummies that indicate if the population of a municipality-

year is above or below a given population size.34 I then run equation (2) with

every placebo treatment (so I run 190 regressions per outcome variable). If the

effect of direct democracy has an actual causal effect, then the estimate of direct

democracy on policy, based on the 100-inhabitant threshold, should be an outlier in

the distribution of placebo coefficients.

I show the results from these tests in Figure 5. I show the empirical cumulative

distribution function of point estimates and t-statistics for the 190 regressions con-

sidered for each variable. I also show the implied p-values, which are the share of

placebo regressions in which I obtain a point estimate (or t-statistic) that is larger

in absolute value than the one for the true threshold. For expenditures, the im-

plied p-value is 0.036 for point estimates, and 0.01 for t-statistics. For revenues, the

p-values are 0.126 and 0.068, respectively and, for deficits, 0.405 and 0.242.

These tests specifically address the concern that the bandwidths used in the main

specifications are inadequate and lead to artificially significant effects.35 However,

if that were the case, then we should see similarly large effects at other thresholds.

These tests show that this is not the case.

34Below 30 (above 220) inhabitants the effects become very imprecisely estimated as there arefew observations below (above) that population size—as explained in Section 4, at 250 inhabitantsthere is another threshold, so I focus on population sizes below that threshold to avoid confoundingeffects.

35For example, it could be argued that the bandwidths used are too large. Even though I calculatethe optimal bandwidth following the standard procedure by Imbens and Kalyanaraman (2012) andshow the results at different fractions of that bandwidth, precision considerations might lead theprocedure to select bandwidths that are too large.


7.5 Donut Regressions

This section considers donut regressions in the spirit of Barreca, Guldi, Lindo, and

Waddell (2011). These regressions exclude from the estimation observations that

are very close to the threshold, where most of the sorting is concentrated. Given

the nature of the sorting process, it is likely that most of the self-selection of mu-

nicipalities into representative democracy is concentrated in a small window around

the threshold. It does not seem plausible that the population size of municipalities

that are self-selecting is much larger than what is strictly necessary to be above

the threshold or that municipalities with a population size far below the threshold

attempt to cross it.36 Thus, finding a similar effect when municipalities very close

to the threshold are excluded would reinforce the credibility of the estimates.

For each outcome variable, I consider six regressions. The first one is the bench-

mark regression. The second excludes observations within a 1% interval around the

threshold (that is, municipalities with 99 and 100 inhabitants), the third excludes

those within an interval of 2%, and so on, until the sixth regression, which excludes

5%. Appendix Table A6 shows these results. It is reassuring that, in line with the

baseline specification, all of the point estimates imply a reduction in expenditures

and revenues, ranging from 4 to 7% in expenditures, and from 1 to 4% in revenues,

while the estimates for deficits remain close to zero. The results, however, are for the

most part not statistically significant, due to a considerable increase in the standard

errors. For example, excluding 5 inhabitants yields a point estimate almost identical

to the baseline (-0.072 versus -0.079, respectively), while almost doubling the stan-

dard error (0.049 versus 0.026). Therefore, it is safe to conclude that the baseline

results are not exclusively driven by the observations very close of the threshold, in

which the sorting could be more worrisome.

8 Mechanisms

In this section, I dig into the mechanisms that may drive the results documented

so far. I start by testing an additional prediction of the model outlined in Section

36For example, a municipality that would have “naturally” had 97 inhabitants may well endup having 103 inhabitants if enough individuals are willing to register so that their municipalityis above the threshold. However, it is unlikely that this municipality would end up having 115inhabitants. Likewise, it is unlikely that a municipality with a population far below the thresholdwould alter its population to reach the threshold: If individuals know that the population is farfrom getting the desired system, they will not even attempt to reach it.


2. Then, I examine four other potential mechanisms, and discuss their plausibility

in light of several additional results and the political and social circumstances in

Spain.

Effects of Political Parties on Policy

The model presented in Section 2 predicts that, under representative democracy,

the two parties converge to high spending. Therefore, we should not see differences

in spending across parties. Here I test whether that prediction is supported by the

data. Although this, by no means, pretends to be a definite test of the validity of

the model, it would be reassuring to find that policy does not differ across parties.

Conversely, finding that it does would raise concerns about the applicability of the

model to the Spanish case.

To test this, I replicate the analysis in Arenas and Bagues (2015) and consider

(representative-democracy) elections in which the right-wing PP and the left-wing

PSOE were the two most-voted parties. I perform a fuzzy regression discontinuity

design, in which the running variable is the difference in the percentage of votes for

the PP and the PSOE (RW Difference) and the treatment variable is having a PP

mayor (RW Mayor).37 Thus, when the running variable is positive (negative), the

most-voted party (RW Winner) is the PP (PSOE). More formally, I estimate:

Outcomemyt = δ + κRWMayormt + g(RWDifferencemt) + εmyt,

RWMayormyt = δ + κRWWinnermt + g(RWDifferencemt) + εmyt, (4)

where the coefficient of interest is κ. I estimate this equation by local linear re-

gression, and I cluster standard errors at the municipality level. Tests for covariate

balance and manipulation of the running variable RW Difference, shown in Table

A10 and Figure A6 in the Appendix, provide assurance that this is a valid empirical

approach.

The first stage is strong. As shown in Panel A of Appendix Table A9, being the

most-voted party increases the probability of obtaining the mayor by 55 percentage

points.38 Panels B to D in Appendix Table A9 show the estimated coefficients of

37One caveat to this approach is that, to obtain precise estimates, I need to use all municipalitiesunder representative democracy, not only those close to the threshold as in the main results.

38There are two forces behind this strong first stage: one is that the most voted party is more

likely to have one more seat in the council, thus normally increasing its bargaining power to appointthe mayor; the second is that, even in cases in which the two most voted parties tie in seats, themost voted party is—surpisingly— much more likely to appoint the mayor. This phenomenon isstudied in depth by Fujiwara and Sanz (2016).


interest κ, and Appendix Figure A5 provides a graphical representation of the first-

stage and reduced-form results. There is no evidence of differences in expenditures

and revenues between PP and PSOE mayors, as the effects are close to zero and

mostly insignificant. For deficits, the results seem to suggest that right-wing mayors

may reduce it, but the effects become insignificant with alternative specifications.39

These results suggest that the increased spending in representative democracy

comes from both parties, as predicted by the model. A similar analysis can be

done for direct-democracy municipalities, in which there is also a mayor. In the

model, political parties play no role in direct democracy, so spending and revenues

should not depend on the affiliation of the mayor, just as we showed it is the case

under representative democracy. The results, shown in Appendix Table A11, reveal

no significant differences in policy across parties.40 However, the results are very

imprecise due to the lack of observations with sufficiently close elections in direct

democracy.

Alternative Mechanisms

(a) Direct democracy may be more prone to elite capture, as was recently proposed

by Hinnerich and Pettersson-Lidbom (2014). An elite who prefers low spending

may exert more influence under direct democracy than in a representative democracy

setting. Hinnerich and Pettersson-Lidbom (2014) provide three arguments that may

show that this was a likely situation in the case studied (Sweden at the beginning

of the 20th century). First, the lack of political parties in direct democracy made it

more difficult for the citizens to solve their collective action problems (e.g., Acemoglu

and Robinson (2008)). Second, the chairman of the town meeting, often a member

of the elite, had substantial power to set the agenda. Third, many decisions at

meetings were taken by an open vote, and there was the potential for intimidation

by the elite.

Although Hinnerich and Pettersson-Lidbom (2014) present compelling evidence

in favor of this mechanism in the case that they study, there are several reasons to

39These results replicate the findings by Arenas and Bagues (2015) for Spain and are in line withthe findings of Ferreira and Gyourko (2009) for United States cities.

40Given that these are first-past-the-post elections, where the most voted candidate automaticallybecomes mayor, this is pure (not fuzzy) regression discontinuity.


doubt that it is relevant in the setting explored in this paper. First, their argument

is based on a context with a conflict between the landed local elite (e.g., farmers)

and citizens (e.g., the agricultural workers) at the time when Sweden was still a

poor, mostly agrarian society. Thus, the landed elite would like to keep the old

labor-repressive economic system, while the citizens would like to have an economic

system based on wage labor. This type of situation is not at all present in contem-

porary Spain. Second, it is not obvious how the elite could capture the meetings in

the Spanish setting. The first two reasons that Hinnerich and Pettersson-Lidbom

(2014) provided for Sweden do not apply: There are political parties under direct

democracy, and the meetings are mostly deliberative, and, therefore, the agenda-

setting power of the mayor is limited. Finally, although it is the case that many

decisions are taken by an open vote and the possibility of intimidation cannot be

completely ruled out, conversations with mayors and other local officials make me

think that intimidation is not an issue in Spanish town meetings. Third, it is not

obvious why the elite would want a smaller local government. Gobernado Rebaque

(2003) shows that fiscal policy is only very slightly progressive at the local level;

even though some expenditures are progressive, this is almost totally offset by the

regresiveness of local taxes.41 Alternatively, if we think of the elite that would want

to capture the meetings as the special interests of the model, then the elite would

prefer a larger government, which is not consistent with the results.

(b) The differences in policy may be driven by costs of participation. An alter-

native model for direct democracy is developed by Osborne, Rosenthal, and Turner

(2000). In this model, individuals decide whether to participate, at a cost, in a

meeting whose outcome is a compromise among the participants’ favorite positions.

In deciding whether to participate, each person compares the cost of participation

with the impact of his or her presence on the compromise. One prediction of the

model is that the outcome will be extremely volatile: in the presence of even a

small amount of randomness, the equilibrium outcome may vary dramatically.42

41A caveat is that Gobernado Rebaque (2003) studies only large municipalities. However, it islikely that the overall system is even more regressive in smaller towns; while the regresiveness ofthe taxes (mainly the property tax) applies to both, the progressive expenditures singled out byGobernado Rebaque (2003) (for example, public transportation) have a very limited role in smallmunicipalities.

42One source of such randomness is the players’ behavior in a mixed-strategy equilibrium. An-other source of randomness is exogenous, and arises in a model in which each player is preventedfrom attending the meeting with some positive probability.


to Osborne, Rosenthal, and Turner (2000), policy will be more variable than in a

representative-democracy setting.43

To shed light on the plausibility of this mechanism, I estimate the effect of direct

democracy on the variability of policy. I proceed in two steps.44 First, I estimate

Equation (2) and obtain the residuals. Second, I estimate Equation (2) letting the

outcome variable be the absolute value of the residuals obtained in the previous

step. The coefficient for DirDem from this second stage captures the effect of direct

democracy on the variability of the outcome.

The results, shown in Appendix Table A12, suggest that there is no effect of

direct democracy on the variability of expenditures or revenues (or, if anything,

that direct democracy leads to less variability). In the main specification, direct

democracy leads to expenditures that deviate from the predicted values around

four percentage points less than in representative democracy. However, the results

become insignificant and closer to zero with other bandwidths or polynomials. For

revenues and deficits, the results are even more clear in pointing to a null effect.

The results are similar if, instead of the absolute value of the residuals, I consider

the squared residuals, the log of the absolute value of the residuals, or the log of

the squared residuals. Therefore, the evidence indicates that, contrary to Osborne,

Rosenthal, and Turner (2000), policy is not more volatile in direct democracy.

(c) Direct democracy may affect the policy that individuals prefer. For example,

the deliberative nature of the meetings may allow citizens to learn by aggregat-

ing information, or it may make the drawbacks of some public expenditures more

salient.45

Although it is not possible to provide a definitive test for this mechanism, I

conduct two tests that can be helpful in assessing its plausibility. First, if the

effects of direct democracy on policy were driven by citizens’ gradual learning from

participating in town meetings, we should observe that the effects grow over time as

municipalities spend more time under direct democracy. Here I test whether that

43Of course, this depends on the mechanics of the representative-democracy system, but none ofthe standard models would predict such big variability.

44Eggers, Freier, Grembi, and Nannicini (2015) follows a similar procedure.45Thus, we may see different policies even if the median voter theorem holds in both representative

and direct democracy and even if there are no previous differences in policy preferences between thecitizens who participate in direct and representative democracy. Under this scenario, differences inpolicy arise because direct democracy brings about a shift in the median voter’s optimal point.

Therefore, it is reasonable to predict that, if direct democracy behaves according


Outcomemyt = αm + γy + f(Populationmt − 100) + βDirDemmt

+4∑

j=2

[Nj,mtfj(Populationmt − 100) + βjNj,mtDirDemmt

]+ umyt, (5)

where Nj,mt is a dummy variable that indicates how many terms municipality m

at year y has been under direct democracy. For a municipality in its first term,

N1,mt = 1 and Nj,mt = 0 for j �= 1 and, analogously, for a municipality in its second

or third terms. For a municipality in its fourth or longer term, N4,mt = 1 and

Nj,mt = 0 for all j �= 4.46 Therefore, β captures the effect of direct democracy in

the first term under direct democracy, and the βj ’s indicate how the effect varies in

subsequent terms.47

The results are shown in Appendix Table A13. For expenditures, the coefficient

onDirDemmt under the optimal bandwidth is -6.3% and significant at the 10% level,

indicating that direct democracy reduces expenditures in the first term. The results

are also significant for other bandwidths (Columns 2 and 3). For N2,mtDirDemmt ,

N3,mtDirDemmt and N4,mtDirDemmt , the coefficients are not significant, suggesting

that the effect does not vary over time. It is true, however, that, due to the short

length of the panel, the estimates are not very precise, so learning cannot be com-

pletely ruled out. A similar pattern appears in the estimation for revenues. For

deficits, neither the coefficient on DirDemmt nor the interaction terms are signifi-

cant.

Second, I ask the following questions: How does being in direct democracy affect

voter behavior in subsequent elections? Does being in this participatory system lead

to more voter turnout? Does it benefit right or left-wing parties? For example,

if the frequent deliberation in town meetings makes individuals more engaged in

46I consider consecutive terms, so a municipality m that switches out of direct democracy andswitches back into direct democracy at term t has N1,mt = 1 at t. A caveat is that I cannot observewhether a municipality followed direct democracy before the sample period starts, as only munici-palities of fewer than 25 (as opposed to 100) inhabitants were required to follow direct democracybefore 1987. Larger municipalities could choose between the two systems. My understanding,based on conversations with local government officials, is that very few municipalities opted fordirect democracy. Thus, I assume that, for municipalities under direct democracy in the first termt = 1, N1,m1 = 1.

47I cannot rule out the possibility that town meetings affect people’s preferences in shorter periodsof time. For example, attending merely one meeting might make individuals change their mindsabout policy, thus driving the results.

is the case, by considering this equation:


the democratic process, we should expect that voter turnout increases in elections

after a municipality switches into direct democracy. In addition to being important

questions in themselves, they can indirectly shed light on whether being under direct

democracy affects individuals’ preferences.

To answer these questions, I test whether being under direct democracy on

term t has an effect on national elections after the beginning of term t—during

period t itself (VotesRightm,t, VotesLeftm,t, VotesFarLeftm,t, Turnoutm,t), and in

the subsequent period t + 1 (VotesRightm,t+1, VotesLeftm,t+1, VotesFarLeftm,t+1,

Turnoutm,t+1)—by estimating Equation (2) with those variables as outcomes. The

results show that there is no effect on any of the outcomes (see Appendix Table

A14). All the coefficients are very close to zero and not significant at conventional

levels.

(d) The differences may be driven by the amount of transfers received from

upper-level governments. In particular, if municipalities in direct democracy have

more projects financed through these transfers, that could explain why they decide

to set lower taxes and fees and spend less. Although some transfers are determined

as a function of population and cannot jump at the threshold, others can, as they are

given at will by provincial and regional governments.48 If direct-democracy munic-

ipalities are better at “lobbying” for these transfers, or if upper-level governments

prefer to fund direct-democracy municipalities, that could lead to more transfers

under direct democracy. It is not clear why we should expect this to happen, but I

can anyway directly test it in the data, by estimating whether there is an effect of

direct democracy on transfers received. In Appendix Table A15, I show the results

of estimating Equation (2) with the log of Transfers as the dependent variable. All

the estimates are insignificant and very close to zero, indicating that there is no

effect of the treatment on transfers. Therefore, it is unlikely that this mechanism

can be driving the results.

48For example, Curto-Grau, Sole-Olle, and Sorribas-Navarro (2015) show that regional govern-ments give more transfers to municipalities that are politically aligned.


Compared with a standard representative-democracy system, direct democracy re-

duces public spending by almost 8%. Revenues are reduced by a similar amount,

thus leaving budget deficits unchanged.

This paper expands our knowledge of the impact of participatory institutions,

by providing the first quasi-experimental evidence on how town meetings affect

spending and taxation decisions in a contemporary setting. Of course, research

from other settings is needed to get a more complete understanding of this issue.

Finally, this paper opens up an interesting avenue for future work, which is to

use this previously unexplored setting to study how other outcomes—for example,

corruption—are affected by the use of participatory meetings.

9 Conclusion

This paper has provided empirical evidence on the effects of direct democracy on

policy. Using a regression discontinuity design in the context of Spanish local gov-

ernments, I have shown that direct democracy results in a smaller government.


References

Acemoglu, D. and J. A. Robinson (2008). Persistence of power, elites, and insti-

tutions. The American Economic Review 98 (1), 267–293.

Aragones, E. and S. Sanchez-Pages (2009). A theory of participatory democracy

based on the real case of Porto Alegre. European Economic Review 53 (1),

56–72.

Arenas, A. and M. Bagues (2015). ¿Realmente importa el color del

partido que gana unas elecciones locales? Nada es Gratis, Avail-

able online at http://nadaesgratis.es/bagues/realmente–importa–el–color–

del–partido–que–gana–unas–elecciones–locales.

Asatryan, Z. (2016). The indirect effects of direct democracy: Local government

size and non-budgetary voter initiatives in Germany. International Tax and

Public Finance 23 (3), 580–601.

Asatryan, Z., T. Baskaran, T. Grigoriadis, and F. Heinemann (2015). Direct

democracy and local public finances under cooperative federalism. The Scan-

dinavian Journal of Economics, forthcoming .

Barreca, A. I., M. Guldi, J. M. Lindo, and G. R. Waddell (2011). Saving babies?

Revisiting the effect of very low birth weight classification. The Quarterly

Journal of Economics 126 (4), 2117–2123.

Berry, C. R. (2014). Direct versus representative democracy: Reassessing the fis-

cal effects. Unpublished manuscript, Kennedy School of Government, Harvard

University .

Bertrand, M., E. Duflo, and S. Mullainathan (2004). How much should we trust

differences-in-differences estimates? Quarterly Journal of Economics 119 (1),

249–275.

Besley, T. and A. Case (2003). Political institutions and policy choices: Evidence

from the United States. Journal of Economic Literature 41 (1), 7–73.

Besley, T. and S. Coate (2008). Issue unbundling via citizens initiatives. Quarterly

Journal of Political Science 3, 379–397.

Besley, T., R. Pande, and V. Rao (2005). Participatory democracy in action:

Survey evidence from south india. Journal of the European Economic Associ-

ation 3 (2-3), 648–657.

Blais, A. and C. H. Achen (2010). Taking civic duty seriously: Political theory

and voter turnout. Unpublished manuscript, Princeton University .

G ( )


Bordignon, M., T. Nannicini, and G. Tabellini (2016). Moderating political ex-

tremism: Single round vs runoff elections under plurality rule. American Eco-

nomic Review 106 (8), 2349–2370.

Calonico, S., M. D. Cattaneo, and R. Titiunik (2014). Robust nonparametric

confidence intervals for regression-discontinuity designs. Econometrica 82 (6),

2295–2326.

Cameron, A. C., J. B. Gelbach, and D. L. Miller (2011). Robust inference with

multiway clustering. Journal of Business & Economic Statistics 29 (2).

Campa, P. (2011). Gender quotas, female politicians and public expenditures:

Quasi-experimental evidence. Working Paper no. 157, Econpubblica, Univer-

sita Bocconi .

Casas-Arce, P. and A. Saiz (2015). Women and power: Unwilling, ineffective, or

held back? Journal of Political Economy 123 (3), 641–669.

Cellini, S. R., F. Ferreira, and J. Rothstein (2010). The value of school facility

investments: Evidence from a dynamic regression discontinuity design. The

Quarterly Journal of Economics 125 (1), 215–261.

Corbi, R., E. Papaioannou, and P. Surico (2014). Federal transfer multipliers:

Quasi-experimental evidence from Brazil. NBER Working Paper No. 20751 .

Curto-Grau, M., A. Sole-Olle, and P. Sorribas-Navarro (2015). Does electoral

competition curb party favoritism? Unpublished manuscript, University of

Barcelona.

De Haan, J. and J.-E. Sturm (1997). Political and economic determinants of oecd

budget deficits and government expenditures: A reinvestigation. European

Journal of Political Economy 13 (4), 739–750.

Deshpande, M. (2014). Does welfare inhibit success? The long-term effects of re-

moving low-income youth from disability insurance. Unpublished manuscript,

MIT .

Eggers, A., R. Freier, V. Grembi, and T. Nannicini (2015). Regression discontinu-

ity designs based on population thresholds: Pitfalls and solutions. Discussion

Papers of DIW Berlin, 1503 .

Feld, L. P. and J. G. Matsusaka (2003). Budget referendums and government

spending: Evidence from Swiss cantons. Journal of Public Economics 87 (12),

2703–2724.


Ferraz, C. and F. Finan (2009). Motivating politicians: The impacts of monetary

incentives on quality and performance. NBER Working Paper No. 14906 .

Ferreira, F. and J. Gyourko (2009). Do political parties matter? Evidence from

US cities. The Quarterly Journal of Economics 124 (1), 399–422.

Frey, B. S. (1994). Direct democracy: Politico-economic lessons from swiss expe-

rience. The American Economic Review 84 (2), 338–342.

Fujiwara, T. (2015). Voting technology, political responsiveness, and infant health:

evidence from Brazil. Econometrica 83 (2), 423–464.

Fujiwara, T. and C. Sanz (2016). Norms in bargaining: Evidence from government

formation in Spain. Unpublished manuscript, Princeton University .

Funk, P. and C. Gathmann (2011). Does direct democracy reduce the size of

government? New evidence from historical data, 1890–2000. The Economic

Journal 121 (557), 1252–1280.

Gagliarducci, S. and T. Nannicini (2013). Do better paid politicians perform

better? Disentangling incentives from selection. Journal of the European Eco-

nomic Association 11 (2), 369–398.

Gagliarducci, S., T. Nannicini, and P. Naticchioni (2011). Electoral rules and

politicians’ behavior: A micro test. American Economic Journal: Economic

Policy 3, 144–174.

Gelman, A. and G. Imbens (2014). Why high-order polynomials should not be

used in regression discontinuity designs. NBER Working Paper No. 20405 .

Gil-Ruiz Gil-Esparza, C. L. and J. Iglesias Quintana (2007). El gasto publico

en Espana en un contexto descentralizado. Presupuesto y Gasto Publico 47,

185–206.

Gobernado Rebaque, J. I. (2003). Un analisis de incidencia fiscal municipal: Efec-

tos redristributivos de los presupuestos de los presupuestos de los ayuntamien-

tos de Castilla y Leon. Hacienda publica espanola (166), 85–114.

Grembi, V., T. Nannicini, and U. Troiano (2012). Policy responses to fiscal re-

straints: A difference-in-discontinuities design. IZA Discussion Papers 6952 .

Hahn, Jinyong, P. T. and W. V. der Klaauw (2001). Identification and estima-

tion of treatment effects with a regression-discontinuity design. Economet-

rica 69 (1), 201–209.

Halvorsen, R. and R. Palmquist (1980). The interpretation of dummy variables

in semilogarithmic equations. American Economic Review 70 (3), 474–75.


Hinnerich, B. T. and P. Pettersson-Lidbom (2014). Democracy, redistribution,

and political participation: Evidence from Sweden 1919–1938. Economet-

rica 82 (3), 961–993.

Imbens, G. and K. Kalyanaraman (2012). Optimal bandwidth choice for the re-

gression discontinuity estimator. The Review of Economic Studies 79 (3), 933–

959.

Imbens, G. W. and T. Lemieux (2008). Regression discontinuity designs: A guide

to practice. Journal of Econometrics 142 (2), 615–635.

Lee, D. S. and D. Card (2008). Regression discontinuity inference with specifica-

tion error. Journal of Econometrics 142 (2), 655–674.

Lee, D. S. and T. Lemieux (2010). Regression discontinuity designs in economics.

Journal of Economic Literature 48 (2), 281–355.

Lemieux, T. and K. Milligan (2008). Incentive effects of social assistance: A

regression discontinuity approach. Journal of Econometrics 142 (2), 807–828.

Matsusaka, J. G. (1995). Fiscal effects of the voter initiative: Evidence from the

last 30 years. Journal of Political Economy 103 (3), 587–623.

Matsusaka, J. G. (2005). Direct democracy works. Journal of Economic Perspec-

tives 19 (2), 185–206.

Matsusaka, J. G. and F. Palda (1999). Voter turnout: How much can we explain?

Public Choice 98 (3-4), 431–446.

McCrary, J. (2008). Manipulation of the running variable in the regression dis-

continuity design: A density test. Journal of Econometrics 142 (2), 698–714.

Olken, B. A. (2010). Direct democracy and local public goods: Evidence from

a field experiment in Indonesia. American Political Science Review 104 (02),

243–267.

Osborne, M. J., J. S. Rosenthal, and M. A. Turner (2000). Meetings with costly

participation. American Economic Review 90, 927–943.

Persson, T. and G. E. Tabellini (2005). The economic effects of constitutions. MIT

press, Cambridge (Ma).

Pettersson-Lidbom, P. (2012). Does the size of the legislature affect the size of

government? Evidence from two natural experiments. Journal of Public Eco-

nomics 96 (3), 269–278.

Porter, J. (2003). Estimation in the regression discontinuity model. Unpublished

manuscript, Harvard University .


Roubini, N. and J. D. Sachs (1989). Political and economic determinants of budget

deficits in the industrial democracies. European Economic Review 33 (5), 903–

933.

Salanova Alcalde, R. (2009). El concejo abierto en Aragon: la ley 9/2009, de 22 de

diciembre, reguladora de los concejos abiertos. Anuario aragones del gobierno

local (1), 193–231.

Sanz, C. (2015). The effect of electoral systems on voter turnout: Evidence from

a natural experiment. Political Science Research and Methods, forthcoming .

Turner, M. and Q. Weninger (2005). Meetings with costly participation: An em-

pirical analysis. The Review of Economic Studies 72 (1), 247–268.

Vila, J. V. i. (2012). El endeudamiento de los municipios: ¿Una cuestion de

comportamiento polıtico? Presupuesto y Gasto Publico 66, 199–216.


Tables

Table 1: Summary Statistics

mean sd p1 p50 p99 count

Expenditures (e per cap.) 671.0 1321.9 74.5 460.1 3524.6 42162

Revenues (e per cap.) 647.3 1290.3 87.9 449.3 3375.5 42162

Deficit (e per cap.) 23.8 270.8 -272.3 0.0 875.4 42162

RW Mayor 0.5 0.5 0.0 1.0 1.0 42073

LW Mayor 0.2 0.4 0.0 0.0 1.0 42073

RW Winner 0.6 0.5 0.0 1.0 1.0 19807

RW Difference (%) 12.4 41.6 -90.0 13.7 96.2 19807

Votes Right (%) 51.6 21.1 2.1 54.1 92.2 42056

Votes Left (%) 30.6 15.5 0.0 29.7 69.4 42056

Votes Far Left (%) 3.2 3.7 0.0 2.2 17.1 35095

Votes Difference (%) 31.0 21.8 0.0 27.5 87.5 42056

Votes Winner (%) 59.3 13.2 32.7 57.9 92.3 42056

Turnout (%) 78.3 8.8 52.9 79.3 95.5 42056

Mean Age (years) 53.0 6.1 39.3 53.0 67.7 27582

Young (%) 9.2 5.5 0.0 8.9 23.4 27582

Middle-Aged (%) 53.4 9.2 29.2 53.6 74.6 27582

Old (%) 37.3 11.6 12.9 37.0 67.9 27582

Immigrants (%) 2.6 4.8 0.0 0.5 22.4 27569

EU Immigrants (%) 45.4 42.1 0.0 41.7 100.0 14338

Population (inhabitants) 130.9 62.4 19.0 128.0 247.0 42162

The unit of observation is a municipality-year. The sample size is smaller for election variables

because some data are missing from the official files. The sample is even smaller for Votes Far Left

because the United Left party did not run in every election. For RW Winner and RW Difference,

the sample is restricted to elections in which the PP and the PSOE were the two most-voted parties.

For the demographic variables, data are available only from 1996. Also note that, in municipalities

with no immigrants, EU Immigrants is missing by construction. See Appendix C for a precise

definition of the variables.


Table 2: Effect of Direct Democracy on Public Finances

Panel A: Log Expenditures

(1) (2) (3) (4) (5) (6) (7)Log Exp. Log Exp. Log Exp. Log Exp. Log Exp. Log Exp. Log Exp.

DirDem -0.0799∗∗∗ -0.0605∗∗ -0.0897∗∗ -0.0729∗∗ -0.0658∗ -0.111∗∗∗ -0.0491∗

(0.0267) (0.0249) (0.0408) (0.0309) (0.0354) (0.0420) (0.0261)

Observations 11932 17646 5964 42026 42026 42026 11932Municipalities 1102 1405 756 2637 2637 2637 1102Specification Linear Linear Linear Order 3 Order 4 Order 5 LinearBandwidth Optimala 1.5 x Opt. .5 x Opt. Full Full Full Opt.Mun. trends Yes

Panel B: Log Revenues

(1) (2) (3) (4) (5) (6) (7)Log Rev. Log Rev. Log Rev. Log Rev. Log Rev. Log Rev. Log Rev.

DirDem -0.0521∗∗ -0.0637∗∗∗ -0.108∗∗∗ -0.0725∗∗ -0.0694∗∗ -0.118∗∗∗ -0.0542∗∗

(0.0261) (0.0231) (0.0412) (0.0284) (0.0326) (0.0392) (0.0235)

Observations 10625 16074 5542 42111 42111 42111 10625Municipalities 1047 1340 729 2638 2638 2638 1047Specification Linear Linear Linear Order 3 Order 4 Order 5 Linear

Bandwidth Optimalb 1.5 x Opt. .5 x Opt. Full Full Full Opt.Mun. trends Yes

Panel C: Deficit (euros)

(1) (2) (3) (4) (5) (6) (7)Deficit Deficit Deficit Deficit Deficit Deficit Deficit

DirDem -7.877 -5.951 0.242 6.323 6.647 3.186 10.32(8.050) (8.456) (9.387) (15.36) (18.96) (21.10) (17.60)

Observations 34570 41761 20647 42111 42111 42111 11958Municipalities 2239 2620 1556 2638 2638 2638 1102Specification Linear Linear Linear Order 3 Order 4 Order 5 LinearBandwidth Optimalc 1.5 x Opt. .5 x Opt. Full Full Full Opt.Mun. trends Yes

Results from estimating Equation (2) (columns 1 to 6) and Equation (3) (column 7). Each columnis a separate regression with a uniform kernel. All regressions include municipality and year fixedeffects, and the last column also includes municipality-specific trends. Standard errors, clustered atboth municipality and the running variable, are in parentheses. I calculate the optimal bandwidthfollowing Imbens and Kalyanaraman (2012)’s procedure. a Optimal BW = 27, b Optimal BW =25. c Optimal BW = 98. ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.


Table 3: Placebo Tests: Covariate Smoothness around the Threshold

(1) (2) (3) (4)Variable Mean Optimal BW Value Opt. BW Results .5 x Opt. BW Results

Votes Right (%) 51.6 29 -0.0866 0.0524[N = 12642] (0.669) (0.966)

Votes Left (%) 30.6 31 -0.258 -0.382[N = 13506] (0.550) (0.771)

Votes Far Left (%) 3.2 28 0.102 0.0510[N = 10023] (0.271) (0.413)

Votes Difference (%) 31.0 33 0.603 1.284[N = 13904] (1.014) (1.382)

Votes Winner (%) 59.3 34 0.330 0.374[N = 14308] (0.603) (0.825)

Turnout (%) 78.3 30 -0.328 -0.144[N = 13099] (0.440) (0.630)

Mean Age (years) 53.0 34 -0.143 -0.0412[N = 9847] (0.191) (0.246)

Young (%) 9.2 41 -0.0993 -0.367[N = 11428] (0.243) (0.290)

Middle-Aged (%) 53.4 37 0.226 0.523[N = 10396] (0.426) (0.494)

Old (%) 37.3 40 -0.177 -0.225[N = 11191] (0.433) (0.501)

Immigrants (%) 2.6 39 -0.240 -0.523[N = 11185] (0.324) (0.397)

EU Immigrants (%) 45.4 29 -0.189 2.452[N = 4056] (3.836) (4.978)

Column 1 shows the mean of the variables. Column 2 shows the optimal bandwidth value and number ofobservations for a placebo test that estimates the effect of direct democracy on the corresponding variable,according to Equation (2). Columns 3 and 4 show the results for the placebo tests: each column is a separatelocal linear regression with a uniform kernel. Standard errors, clustered at both municipality and the runningvariable, are in parentheses. The optimal bandwidth (BW) is based on Imbens and Kalyanaraman (2012).∗p < 0.10, ∗∗p < 0.05, ∗∗∗p < 0.01.


Table 4: Placebo Tests: Lagged Outcomes

Panel A: Log Expenditures (t-1)


DirDem -0.0192 0.00203 -0.0124 -0.0148 -0.000786 -0.0176 -0.00907(0.0291) (0.0287) (0.0420) (0.0354) (0.0417) (0.0493) (0.0305)

Observations 9433 13679 4738 33486 33486 33486 9433Municipalities 1003 1294 678 2506 2506 2506 1003Specification Linear Linear Linear Order 3 Order 4 Order 5 LinearBandwidth Optimala 1.5 x Opt. .5 x Opt. Full Full Full Optimala

Mun. trends Yes

Panel B: Log Revenues (t-1)


DirDem -0.00816 0.00801 -0.00210 -0.00493 0.0175 -0.0126 -0.00908(0.0261) (0.0257) (0.0376) (0.0322) (0.0377) (0.0446) (0.0270)


Bandwidth Optimalb 1.5 x Opt. .5 x Opt. Full Full Full Optimalb

Mun. trends Yes

Panel C: Deficit (t-1) (euros)


DirDem -5.832 -2.534 -3.448 -7.204 -17.86 -11.41 -0.517(9.638) (10.88) (13.62) (11.91) (13.97) (15.42) (9.515)

Observations 10738 16034 5360 33553 33553 33553 9455Municipalities 1100 1436 727 2507 2507 2507 1003Specification Linear Linear Linear Order 3 Order 4 Order 5 LinearBandwidth Optimalc 1.5 x Opt. .5 x Opt. Full Full Full Optimalc

Mun. trends Yes

Results from estimating Equation (2) (columns 1 to 6) and Equation (3) (column 7). Each columnis a separate regression with a uniform kernel. All regressions include municipality and year fixedeffects, and the last column also includes municipality-specific trends. Standard errors, clustered atboth municipality and the running variable, are in parentheses. I calculate the optimal bandwidthfollowing Imbens and Kalyanaraman (2012)’s procedure. a Optimal BW = 27, b Optimal BW =28. c Optimal BW = 31. ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.


Figures

Figure 1: A town meeting in the municipality of Madarcos, in the Madridregion. Source: documentary “Concejo Abierto”, by Carmen Comadran:http://www.tierravoz.com/concejoabierto/el-corto/.

Figure 2: Histogram of population sizes. An observation is a municipality-year.Bins are 10-inhabitant wide.


(a) Log Expenditures (b) Log Revenues

(c) Deficit

Figure 3: Effect of direct democracy on expenditures, revenues, and deficit. I

estimate Outcomemyt = αm + γy +100+OBW∑

j=100−OBW

δjPopulationj ,mt + umyt, where

Populationj ,mt is a dummy that indicates whether municipality m has population

size j at term t. In the y-axis, I plot the estimated coefficients δj , averaged to4-inhabitant-wide bins. I normalize the coefficients so that the average bin imme-diately to the right of the threshold takes the value of zero. The lines are linearfits on δj , fitted separately for observations above and below the threshold. I usethe observations within the optimal bandwidth for LogExpendituresmyt , so that allgraphs show the same range in the x-axis.


(a) Votes Right (b) Votes Left (c) Votes Far Left

(d) Votes Difference (e) Votes Winner (f) Turnout

(g) Mean Age (h) Young (i) Middle-Aged

(j) Old (k) Immigrants (l) EU Immigrants

Figure 4: Placebo Tests: Effect of direct democracy on political and demographic

variables. I estimate Outcomemyt = αm + γy +100+OBW∑

j=100−OBW

δjPopulationj ,mt + umyt,

where Populationj ,mt is a dummy that indicates whether municipality m has popu-

lation size j at term t. In the y-axis, I plot the estimated coefficients δj , averagedto 4-inhabitant-wide bins. I normalize the coefficients so that the average bin im-mediately to the right of the threshold takes the value of zero. The lines are linearfits on δj , fitted separately for observations above and below the threshold. I usethe observations within the optimal bandwidth for LogExpendituresmyt , so that allgraphs show the same range in the x-axis.


(a) Log Expenditures (b) Log Expenditures

(c) Log Revenues (d) Log Revenues

(e) Deficit (f) Deficit

Figure 5: Placebo tests at other population thresholds. I run Equation (2) at all(fake) population thresholds between 30 and 220 inhabitants, by replacing DD witha dummy that indicates whether population is larger than the given populationthreshold. The graphs show the empirical cumulative distribution functions forthe resulting point estimates and t-statistics. The vertical lines show the pointestimates (or t-statistics) for the effects of direct democracy, obtained at the (true)100-inhabitant threshold. The p-value below each graph shows the share of pointestimates (or t-statistics) that are larger in absolute value than the one for the100-inhabitant threshold.


Supplementary Appendices

Appendix A: Model

There is a continuum of individuals who have to decide on two dimensions of policy.

The first one is a primary policy dimension, which I will call ideological stance,

or, simply, ideology, i. Ideology can be left wing, denoted by i, or right wing,

i, so i ∈ {i, i}. We can think of ideology as a salient policy issue, such as the

progressiveness of policy.49 The second issue is targeted or special-interest spending,

denoted by s. Spending also can take two values, s ∈ {s, s}, where s denotes low

(maybe zero) spending and s denotes high spending.

Individuals’ preferences differ over the two issues. With respect to ideology, indi-

viduals are divided into leftists and rightists, v ∈ {l, r}. A fraction γl of individuals

are leftists and prefer i. The rest, γr = 1 − γl, are rightists and prefer i. Let i∗(v)denote the optimal ideological policy from the perspective of an individual of type

v (hence, i∗(l) = i and i∗(r) = i). Without loss of generality, I assume that leftists

are the majority, thus γl > .5. With respect to special-interest spending, individuals

are divided (independently of their preferences in regard to ideology) into citizens

and special interests, w ∈ {c, x}. A fraction γc of individuals are citizens, c, and

prefer low spending, s. The remaining individuals, γx = 1 − γc, are special inter-

ests and prefer high spending, s. The special interests are those who benefit from

special-interest spending, so they do not necessarily constitute an economic elite—

they could be, for example, public employees who are favored by the government.

Special interests are a minority, thus γx < γr. Let s∗(w) denote the optimal ideo-

logical policy from the perspective of an individual of type w (hence, s∗(c) = s and

s∗(x) = s).

49For example, municipalities can use tax deductions to benefit poor individuals. Even if itmay seem surprising that a main “ideological” issue determines the vote at the local level, thecorrelation between votes to the main right (left)-wing party in local and national elections is .63(.57), suggesting that the determinants of voting are closely related at the two levels.


Individuals’ utilities are given by:

ul,c(i, s) = blc [i = i] + θlc [s = s],

ur,c(i, s) = brc [i = i] + θrc [s = s],

ul,x(i, s) = blx [i = i] + θlx [s = s],

ur,x(i, s) = brx [i = i] + θrx [s = s], (6)

where uv,w denotes the utility of type (v, w). Citizens are more concerned about

ideology, so that blc > θlc and brc > θrc. The special interests, in contrast, care

more about special-interest spending, as they benefit directly from it: blx < θlx and

brx < θrx.

Policy under Representative Democracy

Under representative democracy, policy is delegated to an elected representative,

who is an individual and will always implement his or her preferred policy i∗(v), s∗(w)—there is no possibility of commitment. Candidates in the election are put forward

by two political parties P, denoted A and B, P ∈ {A,B}. Each party is comprised

of member individuals bound together by their views on ideology. All members in

Party A are leftists, and all members in B are rightists. Within each party, however,

there can be any combination of citizens and special interests, so that, even if in the

entire population citizens are a majority, that may or may not be the case within

political parties. Each party selects a candidate that a majority of its members

prefer. Because every individual within a party shares the same preferences toward

ideology, the preferences of the majority in the spending dimension will determine

which candidate the party proposes. Let s∗P denote the preferences of the major-

ity of Party P on spending. Parties are not restricted to proposing a member of

their own party; thus, in principle, they could propose somebody with an opposing

ideology, but that will not happen in equilibrium.

To introduce uncertainty into the election, Besley and Coate (2008) assume that

there are some noise voters, a fraction of whom will vote for A’s candidate according

to the realization of some random variable. To keep the notation simple, I assume

instead that the probability that Party A’s candidate wins the election is given by

the share of individuals who prefer Party A’s candidate over Party B’s candidate.50

50The insights and conclusions are the same with both approaches.


Naturally, an individual prefers Party A’s candidate if the policy that Party A’s

candidate will implement gives him or her more utility than will the policy that

Party B’s candidate will implement, according to (6). More formally, an individual

of type {v, w} faced with candidates (vA, wA) and (vB, wB) will favor Party A’s

candidate if uv,w(i∗(vA), s∗(wA)) > uv,w(i

∗(vB), s∗(wB)), and will favor Party B’s

candidate if uv,w(i∗(vA), s∗(wA)) < uv,w(i

∗(vB), s∗(wB)). If both candidates give

him or her the same utility, then he or she will favor each candidate with probability

1/2.

Party members know the election probabilities associated with different candi-

date pairs and take them into account when selecting candidates. An equilibrium is

a pair of candidates, one for each party, such that each party’s majority members do

not want to deviate from their choice, given the other party’s choice. More formally,

a pair of candidates (vA, wA) and (vB, wB) is an equilibrium if type (l, s∗A) individ-uals prefer a type (vA, wA) candidate to any other type of candidate, given that

Party B is running a type (vB, wB) candidate and, conversely, type (r, s∗B) individu-

als prefer a type (vB, wB) candidate to any other type of candidate, given that Party

A is running a type (vA, wA) candidate. Any equilibrium results in a probability

distribution over outcomes. The policy outcome will be that associated with Party

P’s candidate with a probability equal to the chance that Party P’s candidate wins.

Equilibrium under Representative Democracy

Case 1: For each P, s∗P = s.

In this case, special interests are a majority in both political parties. This repre-

sents a situation in which political parties are captured by special interests. In this

case, an equilibrium exists in which both parties run with a special interest of their

preferred ideological position, and that equilibrium is unique. This is summarized

in Proposition 1.

Proposition 1. If, for each P, s∗P = s, then (vA = l, wA = x) and (vB = r, wB = x)

is the unique equilibrium.

Proof. I need to show that if, for each P , s∗P = s, then (vA = l, wA = x) and

(vB = r, wB = x) is the unique equilibrium.

I first show that (vA = l, wA = x) and (vB = r, wB = x) is an equilibrium.

In the proposed equilibrium, Party A wins the election with probability γl, and

its median voter, who is a leftist special interest, gets utility blx+θlx if that happens.


If Party B wins, the median voter in Party A gets utility θlx. Therefore, the expected

utility of the median voter in Party A, which I will denote by u(A), is u∗(A) =

γl(blx + θlx) + γrθlx = θlx + γlblx in the proposed equilibrium.

Party A has three possible deviations: (a) If it deviates to proposing (vA =

l, wA = c), Party A reduces both the probability of winning (to γcγl) and the utility

in case of a win (to blx), so u(A) = γcγlblx+(1− γcγl)θlx < u∗(A). (b) If it deviatesto (vA = r, wA = c), Party A increases the probability of winning from γl to γc but

at the cost of getting zero utility if that happens (it is proposing its least preferred

policy): u(A) = γxθlx < u∗(A). (c) If it deviates to (vA = r, wA = x), then both

parties propose the same policy and win with 1/2 probability, and u(A) = θlx <

u∗(A).Now consider Party B. In the proposed equilibrium, its median voter, who is

a rightist special interest, has expected utility u∗(B) = γr(brx + θrx) + γlθrx =

θrx + γrbrx. Consider the three possible deviations: (a) If it deviates to proposing

(vB = r, wB = c), u(B) = γcγrbrx + (1 − γcγr)θrx < u∗(B). (b) If it deviates to

(vB = l, wB = c), u(B) = γxθrx < u∗(B). (c) If it deviates to (vB = l, wB = x),

u(B) = θrx < u∗(B). This completes the proof that (vA = l, wA = x) and (vB =

r, wB = x) is an equilibrium.

I now show that the equilibrium (vA = l, wA = x) and (vB = r, wB = x) is

unique. To do so, I will show that (vA = l, wA = c) and (vB = r, wB = c) is

not an equilibrium (it is trivial to show that other possible proposals cannot be an

equilibrium). Under (vA = l, wA = c) and (vB = r, wB = c), u∗(A) = γlblx. If A

deviates to (vA = l, wA = x), u(A) = (γcγl+γx)(blx+ θlx) > u∗(A). Thus, A prefers

to deviate, and (vA = l, wA = c) and (vB = r, wB = c) cannot be an equilibrium.

The intuition is as follows. Party A will not want to switch to a leftist citizen, as

it would lower the probability of winning (from γl to γlγc, as it loses all of the votes

of the special interests), and it would lower the utility in case of a win for the median

individual in the party, which is a special interest. The same happens if Party A

switches to a rightist special interest—in this case, the probability of winning goes

down to 1/2, as both parties propose the exact same types of candidates. Finally, a

“radical” switch to a rightist citizen increases the probability of winning (to γl) but

at the cost of sacrificing the preferred policy in both dimensions, so the utility in

the case of winning is zero. The reasoning for Party B is analogous. Finally, note


that both parties running with a citizen of their preferred ideological type cannot be

an equilibrium, as both parties would want to deviate to a special interest of their

preferred ideological type.

Case 2: For each P, s∗P = s.

In this case, citizens are a majority in political parties, as they are in the entire

population. In this case as well, an equilibrium exists in which both parties run

with a special-interest candidate of their preferred ideological position, under certain

conditions that are indicated in the following proposition.

Proposition 2. (i) If, for each P , s∗P = s, and, for each v,

bvcθvc

>γcγx

,

then (vA = l, wA = x) and (vB = r, wB = x) is an equilibrium.

(ii) If, additionally,blcθlc

>γx + γcγl

γx + γcγl − γl

orbrcθrc

>γx + γcγr

γx + γcγr − γr,

then the equilibrium is unique.

Proof. (i) I need to show that, if for each P , s∗P = s, and for each v,

bvcθvc

>γcγx

,

then (vA = l, wA = x) and (vB = r, wB = x) is an equilibrium.

Under the proposed equilibrium, the median voter in Party A, who is a leftist

citizen, has an expected utility of u∗(A) = γlblc. Consider his or her three possible

deviations: (a) If it deviates to (vA = l, wA = c), Party A increases the utility in

case of a win (to blx + γlc) but at the cost of reducing the probability of winning

(to γcγl). Because blcθlc

> γcγx, by assumption, u(A) = γcγl(blc + γlc) < u∗(A). (b) If

it deviates to (vA = r, wA = c), u(A) = γcθlc. Because blcθlc

> γcγx, by assumption,

u(A) < u∗(A). (c) If it deviates to (vA = r, wA = x), u(A) = 0 < u∗(A). For PartyB, the argument is symmetric.


(ii) I need to show that if, additionally,

blcθlc

>γx + γcγl


orbrcθrc

>γx + γcγr


then the equilibrium is unique.

I will first show that (vA = l, wA = c) and (vB = r, wB = c) is not an equilibrium.

Under (vA = l, wA = c) and (vB = r, wB = c), u∗(A) = γlblc+θlc. Consider the best

deviation possible deviation for A, which is (vA = l, wA = x) (it is trivial to show

that the two other possible deviations are worse). Then, u(A) = (γcγl + γx)blc +

(1− γcγl − γx)θlc. Ifblcθlc

>γx + γcγl

γx + γcγl − γl,

then u(A) < u∗(A), so Party A will prefer to deviate. For Party B, the best possible

deviation is (vB = r, wB = x). By symmetry, if

brcθrc

>γx + γcγr


party B will prefer to deviate. Because it is sufficient that one of the two parties

deviates, ifblcθlc

>γx + γcγl


orbrcθrc

>γx + γcγr


then (vA = l, wA = c) and (vB = r, wB = c) is not an equilibrium.

Assumption (i) rules out that parties want to deviate by running with a citi-

zen instead of a special interest. By proposing a citizen, the median voter in the

parties sacrifices some probability of winning (for Party A, it goes down to γcγl

from γl) but obtains some additional utility if the party wins the election, (γcl).

The assumption guarantees that the first effect dominates the second by requiring

that ideology, relative to spending, is sufficiently important to citizens, given the

distribution of types in the population (or, stated differently, that there are enough

special interests, given the preferences of citizens). Assumption (ii) rules out that

both parties’ choosing a citizen is an equilibrium.


Case 3: s∗P = s for some P, s∗P′ = s for P �= P′.This case is a combination of the other two: Citizens are the majority in one party

but not in the other.

Proposition 3. If s∗l = s, s∗r = s, and blc/θlc > γc/γx, then (vA = l, wA = x) and

(vB = r, wB = x) is the unique equilibrium. If s∗l = s, s∗r = s, and brc/θrc > γc/γx,

then (vA = l, wA = x) and (vB = r, wB = x) is the unique equilibrium.

Proof. The proof of Proposition 3 follows directly from the proof of Proposition 1

and the proof of Proposition 2.

Following the logic of the first two cases, now the requirement that special in-

terests are sufficiently large is required only for the party in which citizens are a

majority.

Policy and Equilibrium under Direct Democracy

Under direct democracy, ideology and spending are voted on separately. The equi-

librium outcomes are therefore i = i and s = s. Intuitively, as issues are unbundled,

the position preferred by the median voter (a leftist in ideology and a citizen in

spending) prevails in both dimensions.

In sum, the model predicts that direct democracy will reduce spending if special

interests have captured political parties, or if they are sufficiently large. In addition,

the model predicts that spending in representative democracy should not depend

on which party is in office: Both parties, A and B, converge to the overspending

position. Finally, it is important to note that the model does not yield a clear-cut

prediction in terms of welfare. Even if direct democracy leads to policy more in line

with the median voter’s preferences, a switch from representative to direct democ-

racy creates winners and losers. Citizens, who are the majority, benefit from direct

democracy as a result of reduced special-interest spending, while special interests

lose. But special interests may feel more intensely about the issue. Hence, a measure

of welfare—for example, utilitarian—could rise or fall.


territories in he Early Middle Ages, when neighbors organized themselves in assem-

blies to decide on the government of villages. Traditionally, municipalities them-

selves decided whether to work under direct or representative democracy. A first

attempt to introduced a population threshold took place in 1924, when a national

law (Estatuto Municipal) imposed the use of direct democracy to all municipalities

with fewer than 500 inhabitants. However, this provision was never enforced. Dur-

ing the Second Spanish Republic, another attempt was made to extend the use of

direct democracy to all municipalities with fewer than 500 inhabitants (1935 Law),

but this attempt never materialized either, as the regime lasted only one more year

before the onset of the Spanish Civil War in 1936. The situation did not change

during Franco’s regime. A 1955 law (Ley de Regimen Local) required the following

of the direct-democracy system only for those municipalities in which that was the

traditional form of government. It was not until after the restoration of democracy

after Franco’s death in 1975 that the rules changed substantially. In 1978, national

law (Ley de Elecciones Locales) required all municipalities whose population in the

election-year was smaller than 25 inhabitants to follow the direct democracy-system.

In 1985, a reform extended the requirement to all municipalities with fewer than

100 inhabitants.51 This regulation was in force until 2011, and is the focus of this

paper. In 2011, the law was changed to eliminate the requirement to follow direct

democracy for municipalities with fewer than 100 inhabitants. The rationale for this

change was that most local politicians preferred to avoid direct democracy (which

is consistent with the sorting around the threshold) for the reasons discussed in

Section 5. Municipalities can still adopt the direct-democracy system, by following

the procedure described in the national law (similar to the one described in footnote

51), but no municipality is now required to follow direct democracy. No official data

regarding the use of direct democracy after 2011 exist yet, but partial data indicate

that it is low (8.3%), consistent with the purpose of the change in the law.

51Municipalities with 100 or more inhabitants could follow a demanding procedure to adoptthe direct-democracy system. Specifically, a majority of the citizens of the municipality had tosign a petition, and two-thirds of the members of the council and the regional government had toapprove. To the best of my knowledge, no municipality ever used this procedure. This implies thatthe regression discontinuity component of the estimation is sharp, as the probability of treatmentjumps from 0 to 1 at the threshold.

Appendix B: Historical Overview

The use of concejo abierto dates from centuries ago (for a detailed historical overview,

see Salanova Alcalde (2009)). In a primitive form, they appeared in the Christian


Appendix C: Definition of Variables

In this Appendix, I provide details on the definition of the variables used in the

paper.

Budget Variables

Expenditures is defined as the sum of all (non-financial) chapters of the expenditures

budget, net of transfers:

Expendituresmyt =

7∑k=1

Ek,myt −R4,myt −R7,myt,

where m denotes municipality, y denotes the year, t denotes the term, Ek is expendi-

tures per capita on Chapter k of the expenditures budget, and Rk denotes revenues

per capita from Chapter k of the revenues budget.

Analogously, Revenues, Transfers, and Deficit are defined as follows:

Revenuesmyt =7∑

k=1

Rk,myt −R4,myt −R7,myt,

Transfersmyt = R4,myt +R7,myt − E4,myt − E7,myt,

Deficitmyt = Expendituresmyt − Revenuesmyt .

Elections Variables

I use the following variables from national elections:

Turnoutmt = 100VotesCastmt

ElectoralCensusmt,

VotesRightmt = 100PPVotesmt

ValidVotesmt,

VotesLeftmt = 100PSOEVotesmt

ValidVotesmt,

VotesFarLeftmt = 100IUVotesmt

ValidVotesmt,


where

VotesCastmt = VotesForCandidatesmt + BlankVotesmt + SpoiltVotesmt,

ValidVotesmt = VotesForCandidatesmt + BlankVotesmt.

That is, Turnout measures the proportion of citizens who cast a vote over the

set of potential voters (the electoral census). There is no voter registration in Spain:

Potential voters are all citizens of Spain and other EU countries as well as countries

under Reciprocity Treaties; older than 18, and not disenfranchised by court order.52

Valid votes include votes for candidates and blank votes but not spoiled votes. I use

this denominator because it is relevant for the allocation of seats, as it is used to

determine whether parties reach the election threshold to get seats (3% in national

elections). Accordingly, it is the one that is normally reported by the media.

Observations for year y are from the Congress elections that take place during the

term t to which year y belongs. Because Congress elections have always alternated

perfectly with local elections, there is only one Congress election per term. For

example, for years 1988-1991, I consider the Congress elections of 1989. As explained

in the text, I lag the variables to use them as placebos.

Demographic Variables

The age distribution data are provided by the National Institute for Statistics in

intervals of five years, up to “85 or more” until 2010 and up to “100 or more” from

2011. To calculate the average age Mean Age, I use the mid-points of those intervals.

For ages 85–100 (for which there is no five-year information until 2011), I calculate

the average age in 2011 (89.01) and use it for the rest of the years. For ages “100

or more,” I assume the average is 102.5 years.

I define the share of young people Young as the share of individuals in the first

four intervals (19 years old or younger), Middle-Aged as the share of individuals in

the next nine intervals (ages 20–64), and Old as the share of individuals in the final

five intervals (ages 65 or older).

52Disenfranchisement is mostly for disability reasons. In 2011, the number of disenfranchisedindividuals was 79,398 (including individuals younger than 18) or approximately 0.18% of thepopulation.


Appendix Tables

Table A1: Number of Municipalities and Switches by Government System

Panel A: Number of Municipalities by Government System

Term DirDem RepDem Total

1987–1991 603 1385 1988

1991–1995 614 1319 1933

1995–1999 695 1387 2082

1999–2003 781 1475 2256

2003–2007 797 1486 2283

2007–2011 827 1510 2337

Panel B: Number of Switches in Government System

Term RepDem → DirDem DirDem → RepDem Total

1st to 2nd 50 13 63

2nd to 3rd 93 21 114

3rd to 4th 94 19 113

4th to 5th 69 38 107

5th to 6th 73 43 116

Total 379 134 513

The data refer to the municipalities used in the estimation—that is, those with

a population of 250 or fewer inhabitants, and with non-missing data for some

year(s) of the term.


Table A2: Effect of Direct Democracy on Public Finances (No Fixed Effects)


(1) (2) (3) (4) (5) (6)

Log Exp. Log Exp. Log Exp. Log Exp. Log Exp. Log Exp.

DirDem -0.187∗∗∗ -0.142∗∗∗ -0.156∗∗∗ -0.168∗∗∗ -0.169∗∗∗ -0.138∗∗∗

(0.0357) (0.0365) (0.0288) (0.0403) (0.0397) (0.0445)

Observations 11975 17691 6001 42078 42078 42078

Municipalities 1145 1450 793 2689 2689 2689

Specification Linear Linear Linear Order 3 Order 4 Order 5

Bandwidth Optimala 1.5 x Opt. .5 x Opt. Full Full Full


(1) (2) (3) (4) (5) (6)

Log Rev. Log Rev. Log Rev. Log Rev. Log Rev. Log Rev.

DirDem -0.152∗∗∗ -0.141∗∗∗ -0.152∗∗∗ -0.171∗∗∗ -0.177∗∗∗ -0.147∗∗∗

(0.0320) (0.0329) (0.0248) (0.0379) (0.0371) (0.0414)

Observations 10662 16117 5578 42162 42162 42162

Municipalities 1084 1383 765 2689 2689 2689


Bandwidth Optimalb 1.5 x Opt. .5 x Opt. Full Full Full


(1) (2) (3) (4) (5) (6)

Deficit Deficit Deficit Deficit Deficit Deficit

DirDem -1.851 0.266 -1.722 -3.571 12.82 -6.372

(7.883) (7.585) (9.366) (15.24) (19.27) (19.50)

Observations 34617 41811 20694 42162 42162 42162

Municipalities 2286 2670 1603 2689 2689 2689


Bandwidth Optimalc 1.5 x Opt. .5 x Opt. Full Full Full

Results from estimating Equation (1). Each column is a separate regression with a uniform

kernel. Standard errors, clustered at both municipality and the running variable, are in

parentheses. I calculate the optimal bandwidth following Imbens and Kalyanaraman

(2012)’s procedure. a Optimal BW = 54, b Optimal BW = 53. c Optimal BW = 122. ∗

p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.


Table A3: Placebo Tests: Covariate Smoothness around the Threshold (No FixedEffects)

(1) (2) (3) (4)

Variable Mean Optimal BW Value Opt. BW Results .5 x Opt. BW Results

Votes Right (%) 51.6 94 0.619 0.359

[N = 12687] (1.684) (2.139)

Votes Left (%) 30.6 76 -2.768∗∗∗ -1.984

[N = 13551] (1.003) (1.257)

Votes Far Left (%) 3.2 77 0.0979 0.0346

[N = 10071] (0.280) (0.330)

Votes Difference (%) 31.0 58 2.203 2.308

[N = 13952] (1.517) (2.031)

Votes Winner (%) 59.3 58 0.508 0.761

[N = 14355] (0.915) (1.279)

Turnout (%) 78.3 55 -0.543 -0.944∗

[N = 13142] (0.482) (0.561)

Mean Age (years) 53.0 104 0.275 0.431

[N = 9910] (0.381) (0.398)

Young (%) 9.2 79 0.140 -0.0464

[N = 11483] (0.333) (0.318)

Middle-Aged (%) 53.4 83 -0.897 -1.440∗

[N = 10450] (0.691) (0.848)

Old (%) 37.3 148 0.764 1.306

[N = 11247] (0.803) (0.931)

Immigrants (%) 2.6 150 -0.206 -0.0970

[N = 11242] (0.275) (0.272)

EU Immigrants (%) 45.4 62 -3.182 3.855

[N = 4161] (4.255) (5.538)

Column 1 shows the mean of the variables. Column 2 shows the optimal bandwidth value and number of

observations for a placebo test that estimates the effect of direct democracy on the corresponding variable,

according to Equation (1). Columns 3 and 4 show the results for the placebo tests: each column is a separate

local linear regression with a uniform kernel. Standard errors, clustered at both municipality and the running

variable, are in parentheses. The optimal bandwidth (BW) is based on Imbens and Kalyanaraman (2012).∗p < 0.10, ∗∗p < 0.05, ∗∗∗p < 0.01.


Table A4: Placebo Tests: Lagged Outcomes (No Fixed Effects)

Panel A: Log Expenditures (t-1)

(1) (2) (3) (4) (5) (6)

Log Exp. Log Exp. Log Exp. Log Exp. Log Exp. Log Exp.

DirDem -0.143∗∗∗ -0.131∗∗∗ -0.0817∗∗ -0.155∗∗∗ -0.127∗∗∗ -0.0803

(0.0395) (0.0408) (0.0379) (0.0471) (0.0480) (0.0529)

Observations 9501 13734 4790 33561 33561 33561

Municipalities 1071 1349 730 2581 2581 2581


Bandwidth Optimala 1.5 x Opt. .5 x Opt. Full Full Full

Panel B: Log Revenues (t-1)

(1) (2) (3) (4) (5) (6)

Log Rev. Log Rev. Log Rev. Log Rev. Log Rev. Log Rev.

DirDem -0.131∗∗∗ -0.108∗∗∗ -0.0675∗ -0.145∗∗∗ -0.108∗∗ -0.0697

(0.0392) (0.0408) (0.0396) (0.0466) (0.0472) (0.0534)

Observations 9524 14086 4805 33628 33628 33628

Municipalities 1072 1370 731 2582 2582 2582


Bandwidth Optimalb 1.5 x Opt. .5 x Opt. Full Full Full

Panel C: Deficit (t-1)

(1) (2) (3) (4) (5) (6)

Deficit Deficit Deficit Deficit Deficit Deficit

DirDem -15.76∗ -14.06∗ -8.795 -17.46 -7.398 -26.68∗

(8.187) (7.905) (11.39) (11.82) (14.34) (15.91)

Observations 10808 16103 5415 33628 33628 33628

Municipalities 1170 1505 782 2582 2582 2582


Bandwidth Optimalc 1.5 x Opt. .5 x Opt. Full Full Full

Results from estimating Equation (1). Each column is a separate regression with a uniform

kernel. Standard errors, clustered at both municipality and the running variable, are in

parentheses. I calculate the optimal bandwidth following Imbens and Kalyanaraman

(2012)’s procedure. a Optimal BW = 54, b Optimal BW = 53. c Optimal BW = 122. ∗

p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.


Table A5: Effect by Switches into and out of Direct Democracy



DirDem -0.0799∗∗∗ -0.0665∗∗ -0.0518∗ -0.0786∗ -0.141∗∗∗ -0.0835∗∗ -0.116(0.0267) (0.0289) (0.0268) (0.0439) (0.0452) (0.0362) (0.0792)

Observations 11932 11347 16987 5520 10379 16025 4664Municipalities 1102 1093 1402 736 1090 1398 696Specification Linear Linear Linear Linear Linear Linear LinearBandwidth Optimala Opt. 1.5 x Opt. .5 x Opt. Opt. 1.5 x Opt. .5 x Opt.Switches All Into Into Into Out of Out of Out of



DirDem -0.0521∗∗ -0.0422 -0.0551∗∗ -0.0991∗∗ -0.0877∗ -0.0968∗∗∗ -0.0729(0.0261) (0.0283) (0.0249) (0.0451) (0.0467) (0.0348) (0.0841)

Observations 10625 10054 15436 5107 9076 14464 4281Municipalities 1047 1034 1337 709 1029 1333 666Specification Linear Linear Linear Linear Linear Linear Linear

Bandwidth Optimalb Opt. 1.5 x Opt. .5 x Opt. Opt. 1.5 x Opt. .5 x Opt.Switches All Into Into Into Out of Out of Out of



DirDem -7.877 -6.187 -3.236 1.468 -17.49∗ -13.77 -5.560(8.050) (8.465) (8.572) (9.726) (9.283) (9.398) (10.20)

Observations 34570 33680 40796 19965 32726 39842 19006Municipalities 2239 2233 2614 1555 2232 2613 1554Specification Linear Linear Linear Linear Linear Linear LinearBandwidth Optimalc Opt. 1.5 x Opt. .5 x Opt. Opt. 1.5 x Opt. .5 x Opt.Switches All Into Into Into Out of Out of Out of

Results from estimating Equation (2). Each column is a separate regression with a uniform kernel.All regressions include municipality and year fixed effects. Columns with switches “Into” (“Outof”) exclude from the sample the municipality-years corresponding to terms in which a munici-pality switched from direct (representative) to representative (direct) democracy. Standard errors,clustered at both municipality and the running variable, are in parentheses. I calculate the opti-mal bandwidth following Imbens and Kalyanaraman (2012)’s procedure. a Optimal BW = 27, b

Optimal BW = 25. c Optimal BW = 98. ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.


Table A6: Donut Regressions


(1) (2) (3) (4) (5) (6)Log Exp. Log Exp. Log Exp. Log Exp. Log Exp. Log Exp.

DirDem -0.0799∗∗∗ -0.0671∗∗ -0.0388 -0.0439 -0.0635 -0.0724(0.0267) (0.0290) (0.0325) (0.0376) (0.0417) (0.0496)

Observations 11932 11484 11017 10553 10152 9679Municipalities 1102 1099 1095 1092 1090 1083Specification Linear Linear Linear Linear Linear LinearBandwidth Optimala Opt. Opt. Opt. Opt. Opt.Excluded 0 1% 2% 3% 4% 5%


(1) (2) (3) (4) (5) (6)Log Rev. Log Rev. Log Rev. Log Rev. Log Rev. Log Rev.

DirDem -0.0521∗∗ -0.0428 -0.0111 -0.0169 -0.0141 -0.0358(0.0261) (0.0288) (0.0324) (0.0375) (0.0437) (0.0534)

Observations 10625 10176 9708 9243 8836 8363Municipalities 1047 1044 1037 1034 1030 1023Specification Linear Linear Linear Linear Linear Linear

Bandwidth Optimalb Opt. Opt. Opt. Opt. Opt.Excluded 0 1% 2% 3% 4% 5%

Panel C: Deficit

(1) (2) (3) (4) (5) (6)Deficit Deficit Deficit Deficit Deficit Deficit

DirDem -7.877 -5.328 -4.899 -3.978 -9.502 -13.50(8.050) (8.586) (9.043) (9.562) (8.459) (8.866)

Observations 34570 34121 33653 33189 32783 32309Municipalities 2239 2238 2238 2237 2236 2234Specification Linear Linear Linear Linear Linear LinearBandwidth Optimalc Opt. Opt. Opt. Opt. Opt.Excluded 0 1% 2% 3% 4% 5%

Results from estimating Equation (2). Each column is a separate regression with a uniformkernel. All regressions include municipality and year fixed effects. Standard errors, clusteredat both municipality and the running variable, are in parentheses. I calculate the optimalbandwidth following Imbens and Kalyanaraman (2012)’s procedure. a Optimal BW = 27, b

Optimal BW = 25, c Optimal BW = 98. ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.


Table A7: Robustness to Top-Coding Outliers


(1) (2) (3) (4) (5) (6)Log Exp. Log Exp. Log Exp. Log Exp. Log Exp. Log Exp.

DirDem -0.0799∗∗∗ -0.0815∗∗∗ -0.0625∗∗∗ -0.0797∗∗∗ -0.0394 -0.0793∗∗

(0.0267) (0.0255) (0.0232) (0.0244) (0.0278) (0.0392)

Observations 11932 11932 17646 14763 8914 5964Municipalities 1102 1102 1405 1263 946 756Bandwidth Optimala Opt. 1.5 x Opt. 1.25 x Opt. .75 x Opt. .5 x Opt.Top-Coding NO YES YES YES YES YES


(1) (2) (3) (4) (5) (6)Log Rev. Log Rev. Log Rev. Log Rev. Log Rev. Log Rev.

DirDem -0.0521∗∗ -0.0534∗∗ -0.0700∗∗∗ -0.0742∗∗∗ -0.0492∗ -0.0987∗∗

(0.0261) (0.0252) (0.0217) (0.0234) (0.0263) (0.0394)

Observations 10625 10625 16074 13555 8064 5542Municipalities 1047 1047 1340 1198 892 729

Bandwidth Optimalb Opt. 1.5 x Opt. 1.25 x Opt. .75 x Opt. .5 x Opt.Top-Coding NO YES YES YES YES YES

Panel C: Deficit

(1) (2) (3) (4) (5) (6)Deficit Deficit Deficit Deficit Deficit Deficit

DirDem -7.877 -4.460 -3.072 -3.619 -4.848 -2.212(8.050) (4.990) (4.848) (4.902) (5.211) (5.935)

Observations 34570 34570 41761 38310 29227 20647Municipalities 2239 2239 2620 2438 1960 1556Bandwidth Optimalc Opt. 1.5 x Opt. 1.25 x Opt. .75 x Opt. .5 x Opt.Top-Coding NO YES YES YES YES YES

Results from estimating Equation (2). The top-coded regressions winsorize the observationswith a dependent-variable value above the top or below the bottom 1% of the observationswithin the bandwidth. Each column is a separate regression with a uniform kernel. All regres-sions include municipality and year fixed effects. Standard errors, clustered at both municipalityand the running variable, are in parentheses. I calculate the optimal bandwidth following Im-bens and Kalyanaraman (2012)’s procedure. a Optimal BW = 27, b Optimal BW = 25, c

Optimal BW = 98. ∗p < 0.10, ∗∗p < 0.05, ∗∗∗p < 0.01.


Table A8: Number of Observations at the Threshold

(1) (2) (3) (4) (5)

Obs. Obs. Obs. Obs. Obs.

DirDem 0.0204 0.0168 0.0143 0.00376 0.0462

(0.0197) (0.0186) (0.0190) (0.0265) (0.0426)

Observations 18164 27188 22636 9480 4916

Municipalities 1255 1596 1437 901 671

Bandwidth Optimala 1.5 x Opt. 1.25 x Opt. .75 x Opt. .25 x Opt.

Results from estimating Equation (2). The dependent variable is a dummy that takes the

value of 1 if data are available for a (potential) observation and 0 if the data are missing. Each

column is a separate regression with a uniform kernel. All regressions include municipality and

year fixed effects. Standard errors, clustered at both municipality and the running variable,

are in parentheses. I calculate the optimal bandwidth following Imbens and Kalyanaraman

(2012)’s procedure. aOptimal BW = 73 inhabitants. ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.


Table A9: Effect of Political Parties on Policy under Representative Democracy

Panel A: First Stage: Effect of RW Winner on RW Mayor(1) (2) (3) (4) (5) (6)

RW Mayor RW Mayor RW Mayor RW Mayor RW Mayor RW MayorRW Winner 0.566∗∗∗ 0.550∗∗∗ 0.631∗∗∗ 0.542∗∗∗ 0.524∗∗∗ 0.568∗∗∗

(0.0315) (0.0225) (0.0569) (0.0225) (0.0181) (0.0329)Observations 18772 27717 9476 18873 27806 9549Municipalities 2845 3471 1843 2946 3560 1916Fixed effects NO NO NO YES YES YESBandwidth Optimala 1.5 x Opt. .5 x Opt. Opt. 1.5 x Opt. .5 x Opt.

Panel B: Effect of RW Mayor on Log Expenditures(1) (2) (3) (4) (5) (6)

Log Exp. Log Exp. Log Exp. Log Exp. Log Exp. Log Exp.RW Mayor 0.0346 0.0217 -0.0210 0.0181 0.0283 0.0220

(0.0542) (0.0423) (0.0757) (0.0273) (0.0207) (0.0418)Observations 33481 46776 17557 33401 46708 17459Municipalities 3869 4478 2827 3789 4410 2729Fixed effects NO NO NO YES YES YES

Bandwidth Optimalb 1.5 x Opt. .5 x Opt. Opt. 1.5 x Opt. .5 x Opt.

Panel C: Effect of RW Mayor on Log Revenues(1) (2) (3) (4) (5) (6)

Log Rev. Log Rev. Log Rev. Log Rev. Log Rev. Log Rev.RW Mayor 0.0580 0.0276 -0.00799 0.0355 0.0110 0.0328

(0.0571) (0.0446) (0.0799) (0.0271) (0.0207) (0.0421)Observations 31070 43917 16186 30984 43848 16090Municipalities 3747 4351 2690 3661 4282 2594Fixed effects NO NO NO YES YES YESBandwidth Optimalc 1.5 x Opt. .5 x Opt. Opt. 1.5 x Opt. .5 x Opt.

Panel D: Effect of RW Mayor on Deficit(1) (2) (3) (4) (5) (6)

Deficit Deficit Deficit Deficit Deficit DeficitRW Mayor -20.86∗∗ -12.27 -30.53∗∗ -12.75 -5.076 -17.45


Bandwidth Optimald 1.5 x Opt. .5 x Opt. Opt. 1.5 x Opt. .5 x Opt.

Results from estimating Equation (4). The sample is restricted to elections where the PP and PSOEwere the two most-voted parties. Each column is a separate local linear regression with a uniform kernel.Standard errors, clustered at the municipality level, are in parentheses. Fixed effects are municipalityand year fixed effects. The optimal bandwidth (BW) is based on Imbens and Kalyanaraman (2012).a Optimal BW = 9 percentage points, b Optimal BW = 17 percentage points, c Optimal BW = 15percentage points, d Optimal BW = 10 percentage points. ∗p < 0.10, ∗∗p < 0.05, ∗∗∗p < 0.01.


Table A10: Placebo Tests: Covariate Smoothness for “Effects of Political Parties onPolicy under Representative Democracy”

(1) (2) (3) (4)Variable Mean Optimal BW Value Opt. BW Results .5 x Opt. BW Results

Votes Right (%) 51.6 11 -0.452 -1.948[N = 24142] (0.981) (1.345)

Votes Left (%) 30.6 10 1.255 2.154[N = 20787] (0.936) (1.321)

Votes Far Left (%) 3.2 9 -0.885∗ -0.987[N = 15221] (0.470) (0.649)

Votes Difference (%) 31.0 12 0.208 0.115[N = 25609] (1.039) (1.422)

Votes Winner (%) 59.3 14 0.772 1.022[N = 29447] (0.601) (0.821)

Turnout (%) 78.3 11 -1.787∗∗ -1.128[N = 22498] (0.744) (0.978)

Mean Age (years) 53.0 12 0.691 1.529∗

[N = 16313] (0.679) (0.889)Young (%) 9.2 13 -0.998∗ -1.796∗∗

[N = 18259] (0.593) (0.785)Middle-Aged (%) 53.4 14 -0.396 -0.794

[N = 19014] (0.625) (0.840)Old (%) 37.3 20 1.557∗ 1.130

[N = 25697] (0.870) (1.182)Immigrants (%) 2.6 16 0.0441 0.460

[N = 21368] (0.637) (0.948)EU Immigrants (%) 45.4 14 3.253 1.357

[N = 16898] (3.259) (4.400)

Column 1 shows the mean of the variables. Column 2 shows the optimal bandwidth value and numberof observations for a placebo test that estimates the effect of RW Mayor on the corresponding variable,according to Equation (4). Columns 3 and 4 show the results for the placebo tests: each column is a separatelocal linear regression with a uniform kernel. Standard errors, clustered at both municipality and the runningvariable, are in parentheses. The optimal bandwidth (BW) is based on Imbens and Kalyanaraman (2012).∗p < 0.10, ∗∗p < 0.05, ∗∗∗p < 0.01.


Table A11: Effect of Political Parties on Policy under Direct Democracy

Panel A: Effect of RW Mayor on Log Expenditures(1) (2) (3) (4) (5) (6)

Log Exp. Log Exp. Log Exp. Log Exp. Log Exp. Log Exp.RW Mayor 0.0610 0.0492 -0.117 -0.0637 -0.0596 -0.169


Bandwidth Optimalb 1.5 x Opt. .5 x Opt. Opt. 1.5 x Opt. .5 x Opt.

Panel B: Effect of RW Mayor on Log Revenues(1) (2) (3) (4) (5) (6)

Log Rev. Log Rev. Log Rev. Log Rev. Log Rev. Log Rev.RW Mayor 0.0596 0.0601 -0.0382 -0.0301 -0.0904 0.0261

(0.136) (0.108) (0.230) (0.111) (0.0818) (0.206)Observations 1980 3004 997 1954 2972 982Municipalities 368 489 235 342 457 220Fixed effects NO NO NO YES YES YESBandwidth Optimalc 1.5 x Opt. .5 x Opt. Opt. 1.5 x Opt. .5 x Opt.

Panel C: Effect of RW Mayor on Deficit(1) (2) (3) (4) (5) (6)

Deficit Deficit Deficit Deficit Deficit DeficitRW Mayor 16.79 13.48 21.91 27.06 14.67 -56.13


Bandwidth Optimald 1.5 x Opt. .5 x Opt. Opt. 1.5 x Opt. .5 x Opt.

Results from estimating Equation (4) for direct-democracy municipalities. The sample is restricted toelections where the PP and PSOE were the two most-voted parties. Each column is a separate local linearregression with a uniform kernel. Standard errors, clustered at the municipality level, are in parentheses.Fixed effects are municipality and year fixed effects. The optimal bandwidth (BW) is based on Imbensand Kalyanaraman (2012). a Optimal BW = 33 percentage points, b Optimal BW = 33 percentagepoints, c Optimal BW = 58 percentage points. ∗p < 0.10, ∗∗p < 0.05, ∗∗∗p < 0.01.


Table A12: Effect of Direct Democracy on the Variability of Policy


(1) (2) (3) (4) (5) (6) (7)|uexp,mt| |uexp,mt| |uexp,mt| |uexp,mt| |uexp,mt| |uexp,mt| |uexp,mt|

DirDem -0.0418∗∗ -0.0300∗ -0.00694 -0.0452 -0.0391 -0.0234 -0.0117(0.0204) (0.0154) (0.0384) (0.0286) (0.0364) (0.0454) (0.0220)

Observations 6333 9349 3130 11932 11932 11932 6333Municipalities 782 975 538 1102 1102 1102 782Specification Linear Linear Linear Order 3 Order 4 Order 5 LinearBandwidth Optimala 1.5 x Opt. .5 x Opt. Full Full Full Optimala

Mun. trends Yes


(1) (2) (3) (4) (5) (6) (7)|urev,mt| |uexp,mt| |urev,mt| |urev,mt| |urev,mt| |urev,mt| |urev,mt|

DirDem -0.0348 -0.0245 -0.00604 -0.0354 -0.0370 -0.0424 0.00512(0.0254) (0.0165) (0.0343) (0.0261) (0.0315) (0.0391) (0.0289)


Bandwidth Optimalb 1.5 x Opt. .5 x Opt. Full Full Full Optimalb

Mun. trends Yes

Panel C: Deficit

(1) (2) (3) (4) (5) (6) (7)|udef,mt| |udef,mt| |udef,mt| |udef,mt| |udef,mt| |udef,mt| |udef,mt|

DirDem -11.08 -12.61∗ -8.971 -16.06 -2.256 -11.43 -4.158(7.592) (7.363) (9.212) (12.00) (14.30) (16.41) (13.91)

Observations 28337 34570 15225 34570 34570 34570 11958Municipalities 1927 2239 1285 2239 2239 2239 1102Specification Linear Linear Linear Order 3 Order 4 Order 5 LinearBandwidth Optimalc 1.5 x Opt. .5 x Opt. Full Full Full Optimalc

Mun. trends Yes

Results from estimating Equation (2) (columns 1 to 6) or Equation (3) (column 7). The dependentvariable is the absolute value of the residuals obtained from previously estimating Equation (2)or Equation (3). Each column is a separate regression with a uniform kernel. All regressionsinclude municipality and year fixed effects, and the last column also includes municipality-specifictrends. Standard errors, clustered at both municipality and the running variable, are in parentheses.I calculate the optimal bandwidth following Imbens and Kalyanaraman (2012)’s procedure. a

Optimal BW = 14, b Optimal BW = 10. c Optimal BW = 71. ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01.


Table

A13:Effects

byNumber

ofTermsunder

DirectDem

ocracy

Pan

elA:Expenditures,

Revenues,andDeficit

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

LogExp.

LogExp.

LogExp.

LogRev.

LogRev.

LogRev.

Deficit

Deficit

Deficit

DirDem

-0.0634∗

-0.0638∗

∗-0.107∗∗

-0.0436

-0.0661∗

∗-0.106∗∗

-3.098

-2.188

4.623

(0.0336)

(0.0291)

(0.0483)

(0.0317)

(0.0267)

(0.0451)

(11.27)

(12.00)

(12.25)

DirDem

xN

2-0.0557

-0.0277

0.0673

-0.0587

-0.0272

0.0436

-12.16

-14.92

-25.82∗

(0.0524)

(0.0377)

(0.0958)

(0.0506)

(0.0373)

(0.0879)

(12.60)

(12.01)

(13.72)

DirDem

xN

3-0.106

-0.0491

-0.0176

-0.0566

-0.0941∗

-0.0679

-22.93

-28.11

-11.95

(0.0830)

(0.0578)

(0.133)

(0.0793)

(0.0571)

(0.132)

(21.79)

(21.65)

(23.02)

DirDem

xN

4-0.0683

0.00848

-0.0240

-0.00572

-0.0170

0.00995

-1.626

-6.934

26.38

(0.104)

(0.0786)

(0.125)

(0.108)

(0.0802)

(0.171)

(17.78)

(18.05)

(26.65)

Observations

11932

17646

5964

10625

16074

5542

34570

41761

20647

Municipalities

1102

1405

756

1047

1340

729

2239

2620

1556

Bandwidth

Optimala

1.5

xOpt.

.5xOpt.

Optimalb

1.5

xOpt.

.5xOpt.

Optimalc

1.5

xOpt.

.5xOpt.

Resultsfrom

estimatingEquation(5).

Each

columnisaseparate

regressionwithauniform

kernel.Allregressionsincludemunicipality

andyearfixed

effects.Standard

errors,clustered

atboth

municipality

andtherunningvariable,are

inparentheses.Icalculate

the

optimalbandwidth

follow

ingIm

ben

sandKalyanaraman(2012)’sprocedure.

aOptimalBW

=27,

bOptimalBW

=25,

cOptimal

BW

=98.∗p<

0.10,∗∗

p<

0.05,∗∗∗p<

0.01.


Table

A14:

Effects

ofDirectDem

ocracy

onSubsequentElectionsBeh

avior

Pan

elA:EffectofDirectDem

ocracy

attonElectionVariablesatt

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Votes

Right

Votes

Right

Votes

Left

VotesLeft

Votes

FarL

eft

Votes

FarL

eft

Turnout

Turnout

DirDem

-0.649

-0.353

0.212

0.389

0.281

-0.297

-0.217

-0.376

(0.724)

(1.028)

(0.585)

(0.807)

(0.244)

(0.400)

(0.521)

(0.674)

Observations

11493

5526

12300

6332

10133

4926

11929

5967

Municipalities

1086

727

1125

780

1111

741

1102

754

Bandwidth

Optimala

.5xOpt.

Optimala

.5xOpt.

Optimalb

.5xOpt.

Optimalb

.5xOpt.

Pan

elB:EffectofDirectDem

ocracy

attonElectionVariablesatt+1

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Votes

Right

Votes

Right

Votes

Left

VotesLeft

Votes

FarL

eft

Votes

FarL

eft

Turnout

Turnout

DirDem

-0.686

-1.834

0.300

0.879

0.532∗

0.464

0.327

-0.117

(0.776)

(1.252)

(0.617)

(0.894)

(0.307)

(0.457)

(0.476)

(0.669)

Observations

9263

4390

12845

6414

9753

5039

10917

5231

Municipalities

979

654

1178

797

1090

742

1064

718

Bandwidth

Optimalc

.5xOpt.

Optimalc

.5xOpt.

Optimald

.5xOpt.

Optimald

.5xOpt.

Each

columnisaseparate

locallinearregressionwithauniform

kernel.Standard

errors,clustered

atboth

municipality

andtherunningvariable,

are

inparentheses.Fixed

effects

are

municipality

andyearfixed

effects.Icalculate

theoptimalbandwidth

follow

ingIm

ben

sandKalyanaraman

(2012)’sprocedure.

aOptimalBW

=150,bOptimalBW

=102,cOptimalBW

=87,dOptimalBW

=47.∗ p

<0.10,∗∗p<

0.05,∗∗∗ p

<0.01.


Table A15: Effect of Direct Democracy on Transfers from Upper-Level Governments

(1) (2) (3) (4) (5) (6) (7)Log Tr. Log Tr. Log Tr. Log Tr. Log Tr. Log Tr. Log Tr.

DirDem 0.0109 -0.0202 0.00613 -0.00143 0.00195 0.0496 -0.0229(0.0489) (0.0435) (0.0656) (0.0577) (0.0662) (0.0768) (0.0499)

Observations 12655 19064 6525 40550 40550 40550 11504Municipalities 1163 1506 803 2630 2630 2630 1096Specification Linear Linear Linear Order 3 Order 4 Order 5 LinearBandwidth Optimala 1.5 x Opt. .5 x Opt. Full Full Full Opt.Mun. trends Yes

Results from estimating Equation (2) (columns 1 to 6) and Equation (3) (column 7). Each columnis a separate regression with a uniform kernel. All regressions include municipality and year fixedeffects, and the last column also includes municipality-specific trends. Standard errors, clustered atboth municipality and the running variable, are in parentheses. I calculate the optimal bandwidthfollowing Imbens and Kalyanaraman (2012)’s procedure. a Optimal BW = 31. ∗ p < 0.10, ∗∗

p < 0.05, ∗∗∗ p < 0.01.

Appendix Figures

e4

e1

e2

e3

e4

e1

budget budget budget budget budget budget

election electionpopulation

fi gures

tt - 1 t + 1

Figure A1: Timeline: each unit represents a year y.


(a) Expenditures (b) Revenues

(c) Deficits

Figure A2: Histograms of expenditures, revenues, and deficits. An observation is amunicipality-year. Observations above the 99th and below the 1st percentiles havebeen excluded to facilitate comprehension (90th and 1st for deficits). Bins are 100-euro wide (5 percentage points for deficit).


(a) Log Expenditures (b) Log Revenues

(c) Deficit

Figure A3: Robustness to Bandwidth Choice. Circles represent the estimated treat-ment effect, using different bandwidth choices (x-axis). Lines represent the 95%confidence interval (standard errors clustered at the municipality level). I reportall possible bandwidths from 10 to 150 inhabitants. (The smaller sample sizes forbandwidths below 10 yield large confidence intervals—larger than .2, or around 20%of expenditures and revenues—for smaller bandwidths.)


(a) Log Expenditures (t-1) (b) Log Revenues (t-1)

(c) Deficit (t-1)

Figure A4: Placebo Tests: Effect of direct democracy on lagged expenditures, rev-

enues, and deficit. I estimate Outcomemyt = αm+γy+100+OBW∑

j=100−OBW

δjPopulationj ,mt+

umyt, where Populationj ,mt is a dummy that indicates whether municipality m has

population size j at term t. In the y-axis, I plot the estimated coefficients δj , av-eraged to 4-inhabitant-wide bins. I normalize the coefficients so that the averagebin immediately to the right of the threshold takes the value of zero. The lines arelinear fits on δj , fitted separately for observations above and below the threshold. Iuse the observations within the optimal bandwidth for LogExpendituresmyt , so thatall graphs show the same range in the x-axis.


(a) Expenditures (b) Expenditures

(c) Revenues (d) Deficit

Figure A5: The effect of political parties on policy under representative democracy,following Equation (4). The x-axis shows the difference in the percent of votesbetween the PP and the PSOE in local elections. The sample is restricted to electionsin which those two parties were the two most-voted parties. The bins are localaverages of the dependent variables at 1-percentage-point intervals. The lines arefits of linear regressions run separately at both sides of the threshold.

Figure A6: Histogram of RW Difference (difference in percentage points between thePP votes and the PSOE votes in local elections). An observation is a municipality-year. Bins are 1-percentage-point wide. The sample is restricted to elections inwhich the PP and the PSOE were the two most-voted parties.

BANCO DE ESPAÑA PUBLICATIONS

WORKING PAPERS

1530 CRISTINA FERNÁNDEZ and PILAR GARCÍA PEREA: The impact of the euro on euro area GDP per capita.

1531 IRMA ALONSO ÁLVAREZ: Institutional drivers of capital fl ows.

1532 PAUL EHLING, MICHAEL GALLMEYER, CHRISTIAN HEYERDAHL-LARSEN and PHILIPP ILLEDITSCH: Disagreement

about infl ation and the yield curve.

1533 GALO NUÑO and BENJAMIN MOLL: Controlling a distribution of heterogeneous agents.

1534 TITO BOERI and JUAN F. JIMENO: The unbearable divergence of unemployment in Europe.

1535 OLYMPIA BOVER: Measuring expectations from household surveys: new results on subjective probabilities of future

house prices.

1536 CRISTINA FERNÁNDEZ, AITOR LACUESTA, JOSÉ MANUEL MONTERO and ALBERTO URTASUN: Heterogeneity

of markups at the fi rm level and changes during the great recession: the case of Spain.

1537 MIGUEL SARMIENTO and JORGE E. GALÁN: The infl uence of risk-taking on bank effi ciency: evidence from Colombia.

1538 ISABEL ARGIMÓN, MICHEL DIETSCH and ÁNGEL ESTRADA: Prudential fi lters, portfolio composition and capital ratios

in European banks.

1539 MARIA M. CAMPOS, DOMENICO DEPALO, EVANGELIA PAPAPETROU, JAVIER J. PÉREZ and ROBERTO RAMOS:

Understanding the public sector pay gap.

1540 ÓSCAR ARCE, SAMUEL HURTADO and CARLOS THOMAS: Policy spillovers and synergies in a monetary union.

1601 CHRISTIAN CASTRO, ÁNGEL ESTRADA and JORGE MARTÍNEZ: The countercyclical capital buffer in Spain:

an analysis of key guiding indicators.

1602 TRINO-MANUEL ÑÍGUEZ and JAVIER PEROTE: Multivariate moments expansion density: application of the dynamic

equicorrelation model.

1603 ALBERTO FUERTES and JOSÉ MARÍA SERENA: How fi rms borrow in international bond markets: securities regulation

and market segmentation.

1604 ENRIQUE ALBEROLA, IVÁN KATARYNIUK, ÁNGEL MELGUIZO and RENÉ OROZCO: Fiscal policy and the cycle

in Latin America: the role of fi nancing conditions and fi scal rules.

1605 ANA LAMO, ENRIQUE MORAL-BENITO and JAVIER J. PÉREZ: Does slack infl uence public and private labour

market interactions?

1606 FRUCTUOSO BORRALLO, IGNACIO HERNANDO and JAVIER VALLÉS: The effects of US unconventional monetary

policies in Latin America.

1607 VINCENZO MERELLA and DANIEL SANTABÁRBARA: Do the rich (really) consume higher-quality goods? Evidence from

international trade data.

1608 CARMEN BROTO and MATÍAS LAMAS: Measuring market liquidity in US fi xed income markets: a new synthetic

indicator.

1609 MANUEL GARCÍA-SANTANA, ENRIQUE MORAL-BENITO, JOSEP PIJOAN-MAS and ROBERTO RAMOS: Growing like

Spain: 1995-2007.

1610 MIGUEL GARCÍA-POSADA and RAQUEL VEGAS: Las reformas de la Ley Concursal durante la Gran Recesión.

1611 LUNA AZAHARA ROMO GONZÁLEZ: The drivers of European banks’ US dollar debt issuance: opportunistic funding

in times of crisis?

1612 CELESTINO GIRÓN, MARTA MORANO, ENRIQUE M. QUILIS, DANIEL SANTABÁRBARA and CARLOS TORREGROSA:

Modelling interest payments for macroeconomic assessment.

1613 ENRIQUE MORAL-BENITO: Growing by learning: fi rm-level evidence on the size-productivity nexus.

1614 JAIME MARTÍNEZ-MARTÍN: Breaking down world trade elasticities: a panel ECM approach.

1615 ALESSANDRO GALESI and OMAR RACHEDI: Structural transformation, services deepening, and the transmission

of monetary policy.

1616 BING XU, ADRIAN VAN RIXTEL and HONGLIN WANG: Do banks extract informational rents through collateral?

1617 MIHÁLY TAMÁS BORSI: Credit contractions and unemployment.

1618 MIHÁLY TAMÁS BORSI: Fiscal multipliers across the credit cycle.

1619 GABRIELE FIORENTINI, ALESSANDRO GALESI and ENRIQUE SENTANA: A spectral EM algorithm for dynamic

factor models.

1620 FRANCISCO MARTÍ and JAVIER J. PÉREZ: Spanish public fi nances through the fi nancial crisis.

1621 ADRIAN VAN RIXTEL, LUNA ROMO GONZÁLEZ and JING YANG: The determinants of long-term debt issuance by

European banks: evidence of two crises.

1622 JAVIER ANDRÉS, ÓSCAR ARCE and CARLOS THOMAS: When fi scal consolidation meets private deleveraging.

1623 CARLOS SANZ: The effect of electoral systems on voter turnout: evidence from a natural experiment.

1624 GALO NUÑO and CARLOS THOMAS: Optimal monetary policy with heterogeneous agents.

1625 MARÍA DOLORES GADEA, ANA GÓMEZ-LOSCOS and ANTONIO MONTAÑÉS: Oil price and economic growth:

a long story?

1626 PAUL DE GRAUWE and EDDIE GERBA: Stock market cycles and supply side dynamics: two worlds, one vision?

1627 RICARDO GIMENO and EVA ORTEGA: The evolution of infl ation expectations in euro area markets.

1628 SUSANA PÁRRAGA RODRÍGUEZ: The dynamic effect of public expenditure shocks in the United States.

1629 SUSANA PÁRRAGA RODRÍGUEZ: The aggregate effects of government incometransfer shocks - EU evidence.

1630 JUAN S. MORA-SANGUINETTI, MARTA MARTÍNEZ-MATUTE and MIGUEL GARCÍA-POSADA: Credit, crisis

and contract enforcement: evidence from the Spanish loan market.

1631 PABLO BURRIEL and ALESSANDRO GALESI: Uncovering the heterogeneous effects of ECB unconventional

monetary policies across euro area countries.

1632 MAR DELGADO TÉLLEZ, VÍCTOR D. LLEDÓ and JAVIER J. PÉREZ: On the determinants of fi scal non-compliance:

an empirical analysis of Spain’s regions.

1633 OMAR RACHEDI: Portfolio rebalancing and asset pricing with heterogeneous inattention.

1634 JUAN DE LUCIO, RAÚL MÍNGUEZ, ASIER MINONDO and FRANCISCO REQUENA: The variation of export prices

across and within fi rms.

1635 JUAN FRANCISCO JIMENO, AITOR LACUESTA, MARTA MARTÍNEZ-MATUTE and ERNESTO VILLANUEVA:

Education, labour market experience and cognitive skills: evidence from PIAAC.

1701 JAVIER ANDRÉS, JAVIER J. PÉREZ and JUAN A. ROJAS: Implicit public debt thresholds: an empirical exercise

for the case of Spain.

1702 LUIS J. ÁLVAREZ: Business cycle estimation with high-pass and band-pass local polynomial regression.

1703 ENRIQUE MORAL-BENITO, PAUL ALLISON and RICHARD WILLIAMS: Dynamic panel data modelling using maximum

likelihood: an alternative to Arellano-Bond.

1704 MIKEL BEDAYO: Creating associations as a substitute for direct bank credit. Evidence from Belgium.

1705 MARÍA DOLORES GADEA-RIVAS, ANA GÓMEZ-LOSCOS and DANILO LEIVA-LEON: The evolution of regional

economic interlinkages in Europe.

1706 ESTEBAN GARCÍA-MIRALLES: The crucial role of social welfare criteria for optimal inheritance taxation.

1707 MÓNICA CORREA-LÓPEZ and RAFAEL DOMÉNECH: Service regulations, input prices and export volumes: evidence

from a panel of manufacturing fi rms.

1708 MARÍA DOLORES GADEA, ANA GÓMEZ-LOSCOS and GABRIEL PÉREZ-QUIRÓS: Dissecting US recoveries.

1709 CARLOS SANZ: Direct democracy and government size: evidence from Spain.

Unidad de Servicios AuxiliaresAlcalá, 48 - 28014 Madrid

E-mail: [email protected]

direct democracy and government size: evidence from spain

Documents